vllm-project/vllm
1
0
Fork 0
mirror of https://github.com/vllm-project/vllm.git synced 2025-12-06 15:04:47 +08:00
Code Issues Packages Projects Releases Wiki Activity

Reapply "Install pre-built xformers-0.0.32.post2 built with pt-2.9.0" (#27768)

Browse Source 
Signed-off-by: Huy Do <huydhn@gmail.com>
This commit is contained in:
Huy Do
2025-10-30 10:22:30 -07:00
committed by GitHub
parent 33a0ea5f32
commit ba33e8830d
2 changed files with 2 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
docker/Dockerfile
View File

@@ -361,13 +361,6 @@ RUN --mount=type=bind,from=build,src=/workspace/dist,target=/vllm-workspace/dist
&& uv pip install --system dist/*.whl --verbose \
--extra-index-url ${PYTORCH_CUDA_INDEX_BASE_URL}/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.')
# TODO (huydhn): Remove this once xformers is released for 2.9.0
RUN --mount=type=cache,target=/root/.cache/uv bash - <<'BASH'
. /etc/environment
export TORCH_CUDA_ARCH_LIST='7.5 8.0+PTX 9.0a'
uv pip install --system --no-build-isolation "git+https://github.com/facebookresearch/xformers@v0.0.32.post2"
BASH
# Install FlashInfer pre-compiled kernel cache and binaries
# https://docs.flashinfer.ai/installation.html
RUN --mount=type=cache,target=/root/.cache/uv \

4
requirements/cuda.txt
View File

@@ -9,7 +9,7 @@ torch==2.9.0
torchaudio==2.9.0
# These must be updated alongside torch
torchvision==0.24.0 # Required for phi3v processor. See https://github.com/pytorch/vision?tab=readme-ov-file#installation for corresponding version
# https://github.com/facebookresearch/xformers/releases/tag/v0.0.32.post1
# xformers==0.0.32.post1; platform_system == 'Linux' and platform_machine == 'x86_64' # Requires PyTorch >= 2.8
# Build from https://github.com/facebookresearch/xformers/releases/tag/v0.0.32.post1
xformers==0.0.33+5d4b92a5.d20251029; platform_system == 'Linux' and platform_machine == 'x86_64' # Requires PyTorch >= 2.9
# FlashInfer should be updated together with the Dockerfile
flashinfer-python==0.4.1