xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-06-07 10:09:08 +08:00
Code Issues Packages Projects Releases Wiki Activity

[BugFix] Fix: ImportError when building on hopper systems (#20513)

Browse Source 
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
This commit is contained in:
Lucas Wilkinson 2025-07-06 00:17:30 -04:00 committed by GitHub
parent 432870829d
commit 40b86aa05e
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
4 changed files with 10 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
.github/CODEOWNERS vendored
View File

@ -16,7 +16,7 @@
/vllm/lora @jeejeelee /vllm/lora @jeejeelee
/vllm/reasoning @aarnphm /vllm/reasoning @aarnphm
/vllm/entrypoints @aarnphm /vllm/entrypoints @aarnphm
CMakeLists.txt @tlrmchlsmth CMakeLists.txt @tlrmchlsmth @LucasWilkinson
# Any change to the VllmConfig changes can have a large user-facing impact, # Any change to the VllmConfig changes can have a large user-facing impact,
# so spam a lot of people # so spam a lot of people

5
csrc/ops.h
View File

@ -239,11 +239,6 @@ void cutlass_moe_mm(
torch::Tensor const& b_strides, torch::Tensor const& c_strides, torch::Tensor const& b_strides, torch::Tensor const& c_strides,
bool per_act_token, bool per_out_ch); bool per_act_token, bool per_out_ch);
void cutlass_blockwise_scaled_grouped_mm(
torch::Tensor& output, const torch::Tensor& a, const torch::Tensor& b,
const torch::Tensor& scales_a, const torch::Tensor& scales_b,
const torch::Tensor& problem_sizes, const torch::Tensor& expert_offsets);
void cutlass_fp4_group_mm( void cutlass_fp4_group_mm(
torch::Tensor& output, const torch::Tensor& a, const torch::Tensor& b, torch::Tensor& output, const torch::Tensor& a, const torch::Tensor& b,
const torch::Tensor& a_blockscale, const torch::Tensor& b_blockscales, const torch::Tensor& a_blockscale, const torch::Tensor& b_blockscales,

9
csrc/quantization/cutlass_w8a8/moe/blockwise_scaled_group_mm_sm100.cu
View File

@ -1,3 +1,5 @@
#include "core/registration.h"
#include <torch/all.h> #include <torch/all.h>
#include <cutlass/arch/arch.h> #include <cutlass/arch/arch.h>
@ -364,4 +366,9 @@ void cutlass_blockwise_scaled_grouped_mm(
TORCH_CHECK(false, "Unsupported output tensor type"); TORCH_CHECK(false, "Unsupported output tensor type");
} }
#endif #endif
} }
TORCH_LIBRARY_IMPL_EXPAND(TORCH_EXTENSION_NAME, CUDA, m) {
m.impl("cutlass_blockwise_scaled_grouped_mm",
&cutlass_blockwise_scaled_grouped_mm);
}

3
csrc/torch_bindings.cpp
View File

@ -399,8 +399,7 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
"Tensor scales_a, Tensor scales_b, " "Tensor scales_a, Tensor scales_b, "
"Tensor problem_sizes, Tensor expert_offsets) -> ()", "Tensor problem_sizes, Tensor expert_offsets) -> ()",
{stride_tag}); {stride_tag});
ops.impl("cutlass_blockwise_scaled_grouped_mm", torch::kCUDA, // conditionally compiled so impl registration is in source file
&cutlass_blockwise_scaled_grouped_mm);
// cutlass nvfp4 block scaled group GEMM // cutlass nvfp4 block scaled group GEMM
ops.def( ops.def(