[AMD] Add silu_and_mul, gelu_and_mul, gelu_tanh_and_mul, and gelu_quick kernels for AMD GPUs (#7135)
Co-authored-by: yiakwy-xpu-ml-framework-team <961186938@qq.com> Co-authored-by: HAI <hixiao@gmail.com>
This commit is contained in:
@@ -36,16 +36,18 @@ def _get_version():
|
||||
operator_namespace = "sgl_kernel"
|
||||
include_dirs = [
|
||||
root / "include",
|
||||
root / "include" / "impl",
|
||||
root / "csrc",
|
||||
]
|
||||
|
||||
sources = [
|
||||
"csrc/allreduce/custom_all_reduce.hip",
|
||||
"csrc/allreduce/quick_all_reduce.cu",
|
||||
"csrc/elementwise/activation.cu",
|
||||
"csrc/moe/moe_align_kernel.cu",
|
||||
"csrc/moe/moe_topk_softmax_kernels.cu",
|
||||
"csrc/torch_extension_rocm.cc",
|
||||
"csrc/speculative/eagle_utils.cu",
|
||||
"csrc/torch_extension_rocm.cc",
|
||||
]
|
||||
|
||||
cxx_flags = ["-O3"]
|
||||
@@ -69,6 +71,7 @@ if amdgpu_target not in ["gfx942", "gfx950"]:
|
||||
)
|
||||
sys.exit(1)
|
||||
|
||||
|
||||
hipcc_flags = [
|
||||
"-DNDEBUG",
|
||||
f"-DOPERATOR_NAMESPACE={operator_namespace}",
|
||||
|
||||
Reference in New Issue
Block a user