skip fp8

CMakeLists.txt

@@ -167,7 +167,7 @@ set(VLLM_EXT_SRC
   "csrc/layernorm_kernels.cu"
   "csrc/quantization/squeezellm/quant_cuda_kernel.cu"
   "csrc/quantization/gptq/q_gemm.cu"
-  "csrc/quantization/fp8/fp8_cuda_kernels.cu"
+  # "csrc/quantization/fp8/fp8_cuda_kernels.cu"
   "csrc/cuda_utils_kernels.cu"
   "csrc/moe_align_block_size_kernels.cu"
   "csrc/pybind.cpp")

cmake/utils.cmake

@@ -119,7 +119,7 @@ function (get_torch_gpu_compiler_flags OUT_GPU_FLAGS GPU_LANG)
 
     list(APPEND GPU_FLAGS
       "-DUSE_ROCM"
-      "-DENABLE_FP8_E4M3"
+      # "-DENABLE_FP8_E4M3"
       "-U__HIP_NO_HALF_CONVERSIONS__"
       "-U__HIP_NO_HALF_OPERATORS__"
       "-fno-gpu-rdc"

setup.py

@@ -356,9 +356,9 @@ def get_vllm_version() -> str:
             version += f"+cu{cuda_version_str}"
     elif _is_hip():
         # Get the HIP version
-        hipcc_version = get_hipcc_rocm_version()
-        if hipcc_version != MAIN_CUDA_VERSION:
-            rocm_version_str = hipcc_version.replace(".", "")[:3]
+        # hipcc_version = get_hipcc_rocm_version()
+        # if hipcc_version != MAIN_CUDA_VERSION:
+        #     rocm_version_str = hipcc_version.replace(".", "")[:3]
         #     version += f"+rocm{rocm_version_str}"
         version = get_version()
     elif _is_neuron():