Handle xformers

Signed-off-by: Huy Do <huydhn@gmail.com>

Author: huydhn

docker/Dockerfile

@@ -363,6 +363,10 @@ 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 2.8.0
+RUN --mount=type=cache,target=/root/.cache/uv \
+    uv pip install --system --no-build-isolation "git+https://github.com/facebookresearch/xformers@v0.0.30"
+
 # If we need to build FlashInfer wheel before its release:
 # $ # Note we remove 7.0 from the arch list compared to the list below, since FlashInfer only supports sm75+
 # $ export TORCH_CUDA_ARCH_LIST='7.5 8.0 8.9 9.0a 10.0a 12.0'

requirements/cuda.txt

@@ -10,5 +10,6 @@ torch==2.8.0
 torchaudio==2.8.0
 # These must be updated alongside torch
 torchvision==0.23.0 # Required for phi3v processor. See https://github.com/pytorch/vision?tab=readme-ov-file#installation for corresponding version
+# TODO (huydhn): Re-enable this once xformers is released for 2.8.0
 # https://github.com/facebookresearch/xformers/releases/tag/v0.0.30
-git+https://github.com/facebookresearch/xformers@v0.0.30; platform_system == 'Linux' and platform_machine == 'x86_64'  # Requires PyTorch >= 2.7
+# git+https://github.com/facebookresearch/xformers@v0.0.30; platform_system == 'Linux' and platform_machine == 'x86_64'  # Requires PyTorch >= 2.7