biondizzle/deepseek-v4-quant
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix: patch vLLM deepseek_v4.py directly in image

Browse Source
This commit is contained in:
biondizzle
2026-05-11 06:09:40 +00:00
parent 7c8c6cd67f
commit c8564caf9d
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
Dockerfile
View File

@@ -18,8 +18,8 @@ ENV PYTHONPATH="/root/DeepGEMM:${PYTHONPATH}"
RUN ln -sf /usr/local/lib/python3.12/dist-packages/nvidia/cu13/lib/libnvrtc.so.13 /usr/local/cuda/lib64/libnvrtc.so
RUN cd /root/DeepGEMM && python3 setup.py build_ext --inplace
# Copy our DeepSeek V4 patch into the image
COPY patches/deepseek_v4.py /defaults/deepseek_v4.py
# Copy our DeepSeek V4 patch over vLLM's model file
COPY patches/deepseek_v4.py /usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/deepseek_v4.py
# Verify everything imports
RUN python3 -c "import deep_gemm; print('DeepGEMM NVFP4 OK')" && \