[Bugfix] Fix DSV3 kernels breaking _C and _moe_C on unsupported arches (#35123)
Signed-off-by: mgoin <mgoin64@gmail.com>
This commit is contained in:
@@ -783,7 +783,6 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
|
|||||||
SRCS "${DSV3_FUSED_A_GEMM_SRC}"
|
SRCS "${DSV3_FUSED_A_GEMM_SRC}"
|
||||||
CUDA_ARCHS "${DSV3_FUSED_A_GEMM_ARCHS}")
|
CUDA_ARCHS "${DSV3_FUSED_A_GEMM_ARCHS}")
|
||||||
list(APPEND VLLM_EXT_SRC ${DSV3_FUSED_A_GEMM_SRC})
|
list(APPEND VLLM_EXT_SRC ${DSV3_FUSED_A_GEMM_SRC})
|
||||||
list(APPEND VLLM_GPU_FLAGS "-DENABLE_DSV3_FUSED_A_GEMM=1")
|
|
||||||
message(STATUS "Building dsv3_fused_a_gemm for archs: ${DSV3_FUSED_A_GEMM_ARCHS}")
|
message(STATUS "Building dsv3_fused_a_gemm for archs: ${DSV3_FUSED_A_GEMM_ARCHS}")
|
||||||
else()
|
else()
|
||||||
message(STATUS "Not building dsv3_fused_a_gemm as no compatible archs found "
|
message(STATUS "Not building dsv3_fused_a_gemm as no compatible archs found "
|
||||||
|
|||||||
@@ -745,3 +745,7 @@ void dsv3_fused_a_gemm(torch::Tensor& output, torch::Tensor const& mat_a,
|
|||||||
stream);
|
stream);
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
|
TORCH_LIBRARY_IMPL_EXPAND(TORCH_EXTENSION_NAME, CUDA, m) {
|
||||||
|
m.impl("dsv3_fused_a_gemm", &dsv3_fused_a_gemm);
|
||||||
|
}
|
||||||
|
|||||||
@@ -20,10 +20,12 @@
|
|||||||
|
|
||||||
#include <ATen/ATen.h>
|
#include <ATen/ATen.h>
|
||||||
#include <ATen/cuda/CUDAContext.h>
|
#include <ATen/cuda/CUDAContext.h>
|
||||||
|
#include <torch/all.h>
|
||||||
|
|
||||||
#include <cuda_bf16.h>
|
#include <cuda_bf16.h>
|
||||||
#include <cuda_runtime.h>
|
#include <cuda_runtime.h>
|
||||||
|
|
||||||
|
#include "core/registration.h"
|
||||||
#include "dsv3_router_gemm_utils.h"
|
#include "dsv3_router_gemm_utils.h"
|
||||||
|
|
||||||
static constexpr int DEFAULT_NUM_EXPERTS = 256;
|
static constexpr int DEFAULT_NUM_EXPERTS = 256;
|
||||||
@@ -161,3 +163,7 @@ void dsv3_router_gemm(at::Tensor& output, // [num_tokens, num_experts]
|
|||||||
}
|
}
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
|
TORCH_LIBRARY_IMPL_EXPAND(TORCH_EXTENSION_NAME, CUDA, m) {
|
||||||
|
m.impl("dsv3_router_gemm", &dsv3_router_gemm);
|
||||||
|
}
|
||||||
|
|||||||
@@ -127,7 +127,7 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, m) {
|
|||||||
|
|
||||||
// DeepSeek V3 optimized router GEMM for SM90+
|
// DeepSeek V3 optimized router GEMM for SM90+
|
||||||
m.def("dsv3_router_gemm(Tensor! output, Tensor mat_a, Tensor mat_b) -> ()");
|
m.def("dsv3_router_gemm(Tensor! output, Tensor mat_a, Tensor mat_b) -> ()");
|
||||||
m.impl("dsv3_router_gemm", torch::kCUDA, &dsv3_router_gemm);
|
// conditionally compiled so impl registration is in source file
|
||||||
#endif
|
#endif
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|||||||
@@ -242,7 +242,7 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
|
|||||||
// DeepSeek V3 fused A GEMM (SM 9.0+, bf16 only, 1-16 tokens).
|
// DeepSeek V3 fused A GEMM (SM 9.0+, bf16 only, 1-16 tokens).
|
||||||
ops.def(
|
ops.def(
|
||||||
"dsv3_fused_a_gemm(Tensor! output, Tensor mat_a, Tensor mat_b) -> ()");
|
"dsv3_fused_a_gemm(Tensor! output, Tensor mat_a, Tensor mat_b) -> ()");
|
||||||
ops.impl("dsv3_fused_a_gemm", torch::kCUDA, &dsv3_fused_a_gemm);
|
// conditionally compiled so impl registration is in source file
|
||||||
|
|
||||||
// Quantized GEMM for AWQ.
|
// Quantized GEMM for AWQ.
|
||||||
ops.def(
|
ops.def(
|
||||||
|
|||||||
Reference in New Issue
Block a user