Skip to content

[Bugfix] Build moe_data for both sm100 and sm90 #20086

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 3 commits into from
Jun 27, 2025
+17 −6
Merged

[Bugfix] Build moe_data for both sm100 and sm90 #20086

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
31a03d9
Fix moe_data to be built for both sm100 and sm90
mgoin Jun 25, 2025
7d8c8ab
Remove dupe moe_data.cu compile
mgoin Jun 26, 2025
5d4c6ac
Add get_cutlass_pplx_moe_mm_data to sm100
mgoin Jun 26, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 12 additions & 2 deletions CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -513,6 +513,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
CUDA_ARCHS "${FP4_ARCHS}")
list(APPEND VLLM_EXT_SRC "${SRCS}")
list(APPEND VLLM_GPU_FLAGS "-DENABLE_NVFP4=1")
list(APPEND VLLM_GPU_FLAGS "-DENABLE_CUTLASS_MOE_SM100=1")
message(STATUS "Building NVFP4 for archs: ${FP4_ARCHS}")
else()
message(STATUS "Not building NVFP4 as no compatible archs were found.")
Expand Down Expand Up @@ -547,8 +548,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
# if it's possible to compile MoE kernels that use its output.
cuda_archs_loose_intersection(SCALED_MM_ARCHS "9.0a" "${CUDA_ARCHS}")
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.3 AND SCALED_MM_ARCHS)
set(SRCS "csrc/quantization/cutlass_w8a8/moe/grouped_mm_c3x.cu"
"csrc/quantization/cutlass_w8a8/moe/moe_data.cu")
set(SRCS "csrc/quantization/cutlass_w8a8/moe/grouped_mm_c3x.cu")
set_gencode_flags_for_srcs(
SRCS "${SRCS}"
CUDA_ARCHS "${SCALED_MM_ARCHS}")
Expand All @@ -566,6 +566,16 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
endif()
endif()

# moe_data.cu is used by all CUTLASS MoE kernels.
cuda_archs_loose_intersection(CUTLASS_MOE_DATA_ARCHS "9.0a;10.0a" "${CUDA_ARCHS}")
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.3 AND CUTLASS_MOE_DATA_ARCHS)
set(SRCS "csrc/quantization/cutlass_w8a8/moe/moe_data.cu")
set_gencode_flags_for_srcs(
SRCS "${SRCS}"
CUDA_ARCHS "${CUTLASS_MOE_DATA_ARCHS}")
list(APPEND VLLM_EXT_SRC "${SRCS}")
endif()

#
# Machete kernels

Expand Down
9 changes: 5 additions & 4 deletions csrc/quantization/cutlass_w8a8/scaled_mm_entry.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -241,7 +241,7 @@ void get_cutlass_moe_mm_data(
// mm to run it for.
int32_t version_num = get_sm_version_num();
#if (defined ENABLE_CUTLASS_MOE_SM90 && ENABLE_CUTLASS_MOE_SM90) || \
(defined ENABLE_SCALED_MM_SM100 && ENABLE_SCALED_MM_SM90)
(defined ENABLE_CUTLASS_MOE_SM100 && ENABLE_CUTLASS_MOE_SM100)
get_cutlass_moe_mm_data_caller(topk_ids, expert_offsets, problem_sizes1,
problem_sizes2, input_permutation,
output_permutation, num_experts, n, k,
Expand All @@ -252,7 +252,7 @@ void get_cutlass_moe_mm_data(
false,
"No compiled get_cutlass_moe_mm_data: no cutlass_scaled_mm kernel for "
"CUDA device capability: ",
version_num, ". Required capability: 90");
version_num, ". Required capability: 90 or 100");
}

void get_cutlass_pplx_moe_mm_data(torch::Tensor& expert_offsets,
Expand All @@ -265,7 +265,8 @@ void get_cutlass_pplx_moe_mm_data(torch::Tensor& expert_offsets,
// This function currently gets compiled only if we have a valid cutlass moe
// mm to run it for.
int32_t version_num = get_sm_version_num();
#if defined ENABLE_CUTLASS_MOE_SM90 && ENABLE_CUTLASS_MOE_SM90
#if (defined ENABLE_CUTLASS_MOE_SM90 && ENABLE_CUTLASS_MOE_SM90) || \
(defined ENABLE_CUTLASS_MOE_SM100 && ENABLE_CUTLASS_MOE_SM100)
get_cutlass_pplx_moe_mm_data_caller(expert_offsets, problem_sizes1,
problem_sizes2, expert_num_tokens,
num_local_experts, padded_m, n, k);
Expand All @@ -275,7 +276,7 @@ void get_cutlass_pplx_moe_mm_data(torch::Tensor& expert_offsets,
false,
"No compiled get_cutlass_pplx_moe_mm_data: no cutlass_scaled_mm kernel "
"for CUDA device capability: ",
version_num, ". Required capability: 90");
version_num, ". Required capability: 90 or 100");
}

void cutlass_scaled_mm_azp(torch::Tensor& c, torch::Tensor const& a,
Expand Down