A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 17, 03:00 UTC
[CI/Build] fix TORCH_CUDA_ARCH_LIST in wheel build (#3801)
youkaichao committed
Apr 2, 2024 at 11:54 UTC
205b94942e97f668c01d8f4b476da36277df4c59
1 file changed
+2
-1
.github/workflows/scripts/build.sh
+2
-1
@@ -15,6 +15,7 @@ $python_executable -m pip install -r requirements.txt
15
export MAX_JOBS=1
16
# Make sure punica is built for the release (for LoRA)
17
export VLLM_INSTALL_PUNICA_KERNELS=1
18
-
18
+# Make sure release wheels are built for the following architectures
19
+export TORCH_CUDA_ARCH_LIST="7.0 7.5 8.0 8.6 8.9 9.0+PTX"
20
# Build
21
$python_executable setup.py bdist_wheel --dist-dir=dist