karylab_agents/vllm
1
0
Fork 0
forked from Karylab-cklius/vllm
Code Pull Requests 2 Actions 1 Packages Activity

[Build] Fix CUDA arch build coverage gaps (#45277)

Browse Source 
Signed-off-by: Shengqi Chen <harry-chen@outlook.com>
Co-authored-by: Xin Li <xinli-sw@users.noreply.github.com>
Co-authored-by: ShawRong <ShawRong@users.noreply.github.com>
Co-authored-by: Change72 <Change72@users.noreply.github.com>
This commit is contained in:
Shengqi Chen
2026-06-13 22:09:20 -07:00
committed by GitHub
co-authored by Xin Li ShawRong Change72
parent 54bbf51668
commit 78e7293bb1
14 changed files with 240 additions and 132 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/workflows/scripts/build.sh
+5 -2
View File
@@ -9,7 +9,7 @@ PATH=${cuda_home}/bin:$PATH
LD_LIBRARY_PATH=${cuda_home}/lib64:$LD_LIBRARY_PATH
# Install requirements
if [ "$(echo $2 | cut -d. -f1)" = "12" ]; then
if [ "$(echo "$2" | cut -d. -f1)" = "12" ]; then
sed -i 's/^nvidia-cutlass-dsl\[cu13\]>=/nvidia-cutlass-dsl>=/' requirements/cuda.txt
fi
$python_executable -m pip install -r requirements/build/cuda.txt -r requirements/cuda.txt
@@ -17,7 +17,10 @@ $python_executable -m pip install -r requirements/build/cuda.txt -r requirements
# Limit the number of parallel jobs to avoid OOM
export MAX_JOBS=1
# Make sure release wheels are built for the following architectures
export TORCH_CUDA_ARCH_LIST="7.5 8.0 8.6 8.9 9.0 10.0 12.0+PTX"
# Do not add +PTX here: vLLM filters torch's top-level PTX flag when it
# converts global gencode flags into per-kernel arch lists. If a specific
# kernel needs PTX, add +PTX to that kernel's CMake arch list instead.
export TORCH_CUDA_ARCH_LIST="7.5 8.0 8.6 8.9 9.0 10.0 12.0"
bash tools/check_repo.sh