nick/comfyui-serverless
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Pin SageAttention to commit that supports GPU-less builds
Some checks failed
Build and Push Docker Image / build (push) Failing after 21m35s
Details

Browse Source 
Recent commits broke TORCH_CUDA_ARCH_LIST support, requiring a GPU
during build. Pin to 2aecfa8 which respects the env var.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
Debian
2026-01-03 20:30:17 +00:00
parent c61aca4074
commit 352128aa39
1 changed files with 4 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
Dockerfile
View File

@@ -62,15 +62,16 @@ RUN pip install -r /tmp/requirements.txt && rm -rf /root/.cache/pip
# Compile SageAttention 2.2 from source with no build isolation # Compile SageAttention 2.2 from source with no build isolation
WORKDIR /tmp WORKDIR /tmp
ENV EXT_PARALLEL=1 ENV EXT_PARALLEL=2
ENV NVCC_APPEND_FLAGS="--threads 1" ENV NVCC_APPEND_FLAGS="--threads 2"
ENV MAX_JOBS=2 ENV MAX_JOBS=4
# Target RunPod GPU architectures: # Target RunPod GPU architectures:
# 8.0 = A100, 8.6 = A10/RTX 3090, 8.9 = RTX 4090/L40, 9.0 = H100/H200 # 8.0 = A100, 8.6 = A10/RTX 3090, 8.9 = RTX 4090/L40, 9.0 = H100/H200
# Note: Blackwell (10.0) not yet supported by SageAttention # Note: Blackwell (10.0) not yet supported by SageAttention
ENV TORCH_CUDA_ARCH_LIST="8.0;8.6;8.9;9.0" ENV TORCH_CUDA_ARCH_LIST="8.0;8.6;8.9;9.0"
RUN git clone https://github.com/thu-ml/SageAttention.git && \ RUN git clone https://github.com/thu-ml/SageAttention.git && \
cd SageAttention && \ cd SageAttention && \
git checkout 2aecfa8 && \
pip install --no-build-isolation . && \ pip install --no-build-isolation . && \
cd / && rm -rf /tmp/SageAttention /root/.cache/pip cd / && rm -rf /tmp/SageAttention /root/.cache/pip