init

ByteMLPerf/byte_micro_perf/workloads/gelu.json

+{
+  "operator": "gelu",
+  "iterations": 100,
+  "input_shape_groups": {
+    "inputs": [
+      [
+        [4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072],
+        [8192]
+      ]
+    ]
+  },
+  "dtype": [
+    "float32",
+    "bfloat16",
+    "float16"
+  ]
+}
\ No newline at end of file

ByteMLPerf/byte_micro_perf/workloads/gemm.json

+{
+  "operator": "gemm",
+  "iterations": 100,
+  "input_shape_groups": {
+    "M": [4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072],
+    "KN": [
+      [1024, 1024], 
+      [4096, 4096], 
+      [8192, 8192], 
+      [12288, 12288], 
+      
+      [16384, 32], 
+      [16384, 128], 
+      [16384, 1024], 
+      
+      [32, 16384], 
+      [128, 16384],
+      [1024, 16384] 
+    ]
+  },
+  "dtype": [
+    "float32",
+    "bfloat16",
+    "float16", 
+    "int8"
+  ]
+}
\ No newline at end of file

ByteMLPerf/byte_micro_perf/workloads/gemv.json

+{
+  "operator": "gemv",
+  "iterations": 100,
+  "input_shape_groups": [
+    {
+      "M": [1],
+      "K": [16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192],
+      "N": [4096, 8192]
+    },
+    {
+      "M": [1],
+      "K": [4096, 8192],
+      "N": [16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192]
+    }
+  ],
+  "dtype": [
+    "float32",
+    "bfloat16",
+    "float16", 
+    "int8"
+  ]
+}
\ No newline at end of file

ByteMLPerf/byte_micro_perf/workloads/group_gemm.json

+{
+  "operator": "group_gemm", 
+  "iterations": 100, 
+  "input_shape_groups": {
+    "gemm_group": [1, 2, 3, 4, 5, 6, 7, 8],
+    "batch": [1, 2, 4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192],
+    "KN": [[32, 16384], [16384, 32], [16384, 16384]]
+  },
+  "dtype": [ 
+    "float32", 
+    "bfloat16", 
+    "float16", 
+    "int8"
+  ]
+}
\ No newline at end of file

ByteMLPerf/byte_micro_perf/workloads/host2device.json

+{
+  "operator": "host2device",
+  "iterations": 100,
+  "input_shape_groups": {
+    "inputs": [
+      [
+        [4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072],
+        [1024]
+      ]
+    ]
+  },
+  "dtype": [
+    "float32",
+    "bfloat16",
+    "float16"
+  ]
+}
\ No newline at end of file

ByteMLPerf/byte_micro_perf/workloads/index_add.json

+{
+  "operator": "index_add",
+  "iterations": 100,
+  "input_shape_groups": {
+    "inputs": [
+      [
+        [1024],
+        [1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072, 262144, 524288]
+      ],
+      [
+        [1024],
+        [1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072, 262144, 524288]
+      ]
+    ]
+  },
+  "dtype": [
+    "float32",
+    "bfloat16", 
+    "float16"
+  ]
+}
\ No newline at end of file

ByteMLPerf/byte_micro_perf/workloads/layernorm.json

+{
+  "operator": "layernorm",
+  "iterations": 100,
+  "input_shape_groups": {
+    "inputs": [
+      [
+        [1024],
+        [1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072, 262144, 524288]
+      ]
+    ]
+  },
+  "dtype": [
+    "float32",
+    "bfloat16",
+    "float16"
+  ]
+}
\ No newline at end of file

ByteMLPerf/byte_micro_perf/workloads/log.json

+{
+  "operator": "log",
+  "iterations": 100,
+  "input_shape_groups": {
+    "inputs": [
+      [
+        [4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072],
+        [8192]
+      ]
+    ]
+  },
+  "dtype": [
+    "float32",
+    "bfloat16",
+    "float16"
+  ]
+}
\ No newline at end of file

ByteMLPerf/byte_micro_perf/workloads/mul.json

+{
+  "operator": "mul",
+  "iterations": 100,
+  "input_shape_groups": {
+    "inputs": [
+      [
+        [4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072],
+        [8192]
+      ],
+      
+      [
+        [4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072],
+        [8192]
+      ]
+    ]
+  },
+  "dtype": [
+    "float32",
+    "bfloat16",
+    "float16"
+  ]
+}
\ No newline at end of file

ByteMLPerf/byte_micro_perf/workloads/p2p.json

+{
+  "operator": "p2p",
+  "iterations": 100,
+  "input_shape_groups": {
+    "inputs": [
+      [
+        [8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072, 262144, 524288, 1048576, 2097152],
+        [1024]
+      ],
+      [
+        [8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072, 262144, 524288, 1048576, 2097152],
+        [1024]
+      ]
+    ]
+  },
+  "dtype": [
+    "float32",
+    "bfloat16",
+    "float16"
+  ],
+  "group": [
+    2,
+    4,
+    8
+  ]
+}
\ No newline at end of file

ByteMLPerf/byte_micro_perf/workloads/reduce_max.json

+{
+  "operator": "reduce_max",
+  "iterations": 100,
+  "input_shape_groups": {
+    "inputs": [
+      [
+        [1024],
+        [1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072, 262144, 524288]
+      ]
+    ]
+  },
+  "dtype": [
+    "float32",
+    "bfloat16",
+    "float16"
+  ]
+}
\ No newline at end of file

ByteMLPerf/byte_micro_perf/workloads/reduce_min.json

+{
+  "operator": "reduce_min",
+  "iterations": 100,
+  "input_shape_groups": {
+    "inputs": [
+      [
+        [1024],
+        [1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072, 262144, 524288]
+      ]
+    ]
+  },
+  "dtype": [
+    "float32",
+    "bfloat16",
+    "float16"
+  ]
+}
\ No newline at end of file

ByteMLPerf/byte_micro_perf/workloads/reduce_sum.json

+{
+  "operator": "reduce_sum",
+  "iterations": 100,
+  "input_shape_groups": {
+    "inputs": [
+      [
+        [1024],
+        [1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072, 262144, 524288]
+      ]
+    ]
+  },
+  "dtype": [
+    "float32",
+    "bfloat16",
+    "float16"
+  ]
+}
\ No newline at end of file

ByteMLPerf/byte_micro_perf/workloads/reducescatter.json

+{
+  "operator": "reducescatter",
+  "iterations": 100,
+  "input_shape_groups": {
+    "inputs": [
+      [
+        [8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072, 262144, 524288, 1048576, 2097152],
+        [1024]
+      ]
+    ]
+  },
+  "dtype": [
+    "float32",
+    "bfloat16",
+    "float16"
+  ],
+  "group": [
+    2,
+    4,
+    8
+  ]
+}
\ No newline at end of file

ByteMLPerf/byte_micro_perf/workloads/scatter.json

+{
+    "operator": "scatter", 
+    "iterations": 100, 
+    "input_shape_groups": {
+      "inputs": [
+        [
+          [1024], 
+          [1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072, 262144, 524288]
+        ]
+      ]
+    }, 
+    "dtype": [
+      "float32", 
+      "bfloat16", 
+      "float16"
+    ]
+  }
\ No newline at end of file

ByteMLPerf/byte_micro_perf/workloads/silu.json

+{
+  "operator": "silu",
+  "iterations": 100,
+  "input_shape_groups": {
+    "inputs": [
+      [
+        [4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072],
+        [8192]
+      ]
+    ]
+  },
+  "dtype": [
+    "float32",
+    "bfloat16",
+    "float16"
+  ]
+}
\ No newline at end of file

ByteMLPerf/byte_micro_perf/workloads/sin.json

+{
+  "operator": "sin",
+  "iterations": 100,
+  "input_shape_groups": {
+    "inputs": [
+      [
+        [4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072],
+        [8192]
+      ]
+    ]
+  },
+  "dtype": [
+    "float32",
+    "bfloat16",
+    "float16"
+  ]
+}
\ No newline at end of file

ByteMLPerf/byte_micro_perf/workloads/softmax.json

+{
+  "operator": "softmax",
+  "iterations": 100,
+  "input_shape_groups": {
+    "inputs": [
+      [
+        [1024],
+        [1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072, 262144, 524288]
+      ]
+    ]
+  },
+  "dtype": [
+    "float32",
+    "bfloat16",
+    "float16"
+  ]
+}
\ No newline at end of file

ByteMLPerf/byte_micro_perf/workloads/sort.json

+{
+  "operator": "sort",
+  "iterations": 100,
+  "input_shape_groups": {
+    "inputs": [
+      [
+        [1024],
+        [1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072, 262144, 524288]
+      ]
+    ]
+  },
+  "dtype": [
+    "float32",
+    "bfloat16",
+    "float16"
+  ]
+}
\ No newline at end of file

ByteMLPerf/byte_micro_perf/workloads/sqrt.json

+{
+  "operator": "sqrt",
+  "iterations": 100,
+  "input_shape_groups": {
+    "inputs": [
+      [
+        [4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072],
+        [8192]
+      ]
+    ]
+  },
+  "dtype": [
+    "float32",
+    "bfloat16",
+    "float16"
+  ]
+}
\ No newline at end of file