diff --git a/3rdparty/CMakeLists.txt b/3rdparty/CMakeLists.txt index 5bd3a6ee98..59076e14c9 100644 --- a/3rdparty/CMakeLists.txt +++ b/3rdparty/CMakeLists.txt @@ -19,7 +19,7 @@ FetchContent_Declare( FetchContent_Declare( cutlass GIT_REPOSITORY https://github.com/NVIDIA/cutlass - GIT_TAG v4.2.1 # f3fde58372d33e9a5650ba7b80fc48b3b49d40c8 + GIT_TAG v4.3.0 # e67e63c331d6e4b729047c95cf6b92c8454cba89 GIT_SHALLOW TRUE SOURCE_SUBDIR dont-add-this-project-with-add-subdirectory) diff --git a/cpp/tensorrt_llm/cutlass_extensions/include/cutlass_extensions/gemm/collective/builders/sm90_gmma_builder_mixed_input.inl b/cpp/tensorrt_llm/cutlass_extensions/include/cutlass_extensions/gemm/collective/builders/sm90_gmma_builder_mixed_input.inl index e35b42ddf7..aa6fa76924 100644 --- a/cpp/tensorrt_llm/cutlass_extensions/include/cutlass_extensions/gemm/collective/builders/sm90_gmma_builder_mixed_input.inl +++ b/cpp/tensorrt_llm/cutlass_extensions/include/cutlass_extensions/gemm/collective/builders/sm90_gmma_builder_mixed_input.inl @@ -177,13 +177,13 @@ public: static constexpr int PipelineStages = IsMixedInput ? (IsArrayOfPointersGemm ? detail::compute_stage_count_or_override_single_affine_transformed_input(StageCountType{}) + RealElementA, RealElementB, ElementScale, ElementZero, TileShape_MNK, SmemAlignment, + StageCountType::bytes>(StageCountType{}) : detail::compute_stage_count_or_override_single_affine_transformed_input< detail::sm90_smem_capacity_bytes, RealElementA, RealElementB, ElementScale, ElementZero, - TileShape_MNK, StageCountType::bytes, SmemAlignment>(StageCountType{})) + TileShape_MNK, SmemAlignment, StageCountType::bytes>(StageCountType{})) : detail::compute_stage_count_or_override(StageCountType{}); + TileShape_MNK, SmemAlignment, StageCountType::bytes>(StageCountType{}); using DispatchPolicy = cute::conditional_t