bump sgl-kernel 0.0.5.post4 (#4768)

This commit is contained in:
Yineng Zhang
2025-03-28 14:40:53 -07:00
committed by GitHub
parent 2bb0e7cf43
commit 92941ce7b5
5 changed files with 12 additions and 4 deletions

View File

@@ -86,6 +86,7 @@ option(SGL_KERNEL_ENABLE_SM100A "Enable SM100A" OFF)
option(SGL_KERNEL_ENABLE_SM90A "Enable SM90A" OFF)
option(SGL_KERNEL_ENABLE_BF16 "Enable BF16" ON)
option(SGL_KERNEL_ENABLE_FP8 "Enable FP8" ON)
option(SGL_KERNEL_ENABLE_FP4 "Enable FP4" OFF)
if ("${CUDA_VERSION}" VERSION_GREATER_EQUAL "12.8" OR SGL_KERNEL_ENABLE_SM100A)
list(APPEND SGL_KERNEL_CUDA_FLAGS
@@ -118,6 +119,12 @@ if (SGL_KERNEL_ENABLE_FP8)
)
endif()
if (SGL_KERNEL_ENABLE_FP4)
list(APPEND SGL_KERNEL_CUDA_FLAGS
"-DENABLE_NVFP4=1"
)
endif()
string(REPLACE "-D__CUDA_NO_HALF_OPERATORS__" "" CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS}")
string(REPLACE "-D__CUDA_NO_HALF_CONVERSIONS__" "" CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS}")
string(REPLACE "-D__CUDA_NO_BFLOAT16_CONVERSIONS__" "" CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS}")