mirror of
https://git.datalinker.icu/vllm-project/vllm.git
synced 2026-05-27 11:34:25 +08:00
Merge 1133c53c0e6814d07c2754d9d4049b5583ca5ab1 into 254f6b986720c92ddf97fbb1a6a6465da8e87e29
This commit is contained in:
commit
932e928da7
@ -62,6 +62,12 @@ install(CODE "set(CMAKE_INSTALL_PREFIX \"\${CMAKE_INSTALL_PREFIX}/vllm/\")" ALL_
|
|||||||
FetchContent_MakeAvailable(vllm-flash-attn)
|
FetchContent_MakeAvailable(vllm-flash-attn)
|
||||||
message(STATUS "vllm-flash-attn is available at ${vllm-flash-attn_SOURCE_DIR}")
|
message(STATUS "vllm-flash-attn is available at ${vllm-flash-attn_SOURCE_DIR}")
|
||||||
|
|
||||||
|
# Suppress ptxas warnings for flash attention compilation
|
||||||
|
if(VLLM_GPU_LANG STREQUAL "CUDA")
|
||||||
|
# Add ptxas flags to suppress C7520 warnings about wgmma.mma_async serialization
|
||||||
|
set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -Xptxas --diag_suppress=7520")
|
||||||
|
endif()
|
||||||
|
|
||||||
# Restore the install prefix
|
# Restore the install prefix
|
||||||
install(CODE "set(CMAKE_INSTALL_PREFIX \"\${OLD_CMAKE_INSTALL_PREFIX}\")" ALL_COMPONENTS)
|
install(CODE "set(CMAKE_INSTALL_PREFIX \"\${OLD_CMAKE_INSTALL_PREFIX}\")" ALL_COMPONENTS)
|
||||||
install(CODE "set(CMAKE_INSTALL_LOCAL_ONLY TRUE)" ALL_COMPONENTS)
|
install(CODE "set(CMAKE_INSTALL_LOCAL_ONLY TRUE)" ALL_COMPONENTS)
|
||||||
|
|||||||
Loading…
x
Reference in New Issue
Block a user