kanshan/TensorRT-LLMs
1
0
Fork 0
mirror of https://github.com/NVIDIA/TensorRT-LLM.git synced 2026-02-01 00:31:24 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

fix cutlass moe not falling back

Browse Source 
Signed-off-by: Xiwen Yu <13230610+VALLIS-NERIA@users.noreply.github.com>
This commit is contained in:
Xiwen Yu 2025-08-30 14:50:54 +08:00
parent c425c1214d
commit 0fb835d7c2
2 changed files with 4 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
3rdparty/cutlass vendored

@ -1 +1 @@
Subproject commit a49a78ffefc86a87160dfe0ccc3a3a2d1622c918
Subproject commit b2dd65dc864e09688245b316ac46c4a6cd07e15c

3
cpp/tensorrt_llm/kernels/cutlass_kernels/moe_gemm/moe_gemm_template_dispatch_tma_ws.h
View File

@ -113,6 +113,9 @@ void dispatchMoeGemmSelectBiasTmaWarpSpecialized(TmaWarpSpecializedGroupedGemmIn
"passing 103-real as an arch to build_wheel.py.");
first_time = false;
}
return dispatchMoeGemmSelectBiasTmaWarpSpecialized<cutlass::arch::Sm100, T, WeightType, OutputType, EpilogueTag,
FUSION, TileShape, ClusterShape>(
hopper_input, num_experts, multi_processor_count, stream, occupancy, workspace_size);
}
#endif
#ifndef COMPILE_BLACKWELL_TMA_GROUPED_GEMMS