skip triton_kernels

CMakeLists.txt

@@ -1090,7 +1090,7 @@ endif()
 ]]
 
 # For CUDA and HIP builds also build the triton_kernels external package.
-if(VLLM_GPU_LANG STREQUAL "CUDA" OR VLLM_GPU_LANG STREQUAL "HIP")
+if(VLLM_GPU_LANG STREQUAL "CUDA")
     include(cmake/external_projects/triton_kernels.cmake)
 endif()
 

setup.py

@@ -292,7 +292,7 @@ class cmake_build_ext(build_ext):
             os.makedirs(os.path.dirname(dst_file), exist_ok=True)
             self.copy_file(file, dst_file)
 
-        if _is_cuda() or _is_hip():
+        if _is_cuda():
             # copy vllm/third_party/triton_kernels/**/*.py from self.build_lib
             # to current directory so that they can be included in the editable
             # build
@@ -727,7 +727,7 @@ if _is_cuda() or _is_hip():
     ext_modules.append(CMakeExtension(name="vllm.cumem_allocator"))
     # Optional since this doesn't get built (produce an .so file). This is just
     # copying the relevant .py files from the source repository.
-    ext_modules.append(CMakeExtension(name="vllm.triton_kernels", optional=True))
+    # ext_modules.append(CMakeExtension(name="vllm.triton_kernels", optional=True))
 
 # if _is_hip():
 #     ext_modules.append(CMakeExtension(name="vllm._rocm_C"))