gemma4-inference / Dockerfile
yatin-superintelligence's picture
Upload Dockerfile with huggingface_hub
3eb3e17 verified
Raw
History Blame Contribute Delete
357 Bytes
FROM vllm/vllm-openai:latest
# Gemma 4 (model_type=gemma4) needs newer transformers than vLLM bundles.
# TGI does NOT support gemma4 yet — this image uses vLLM which does.
RUN pip install --no-cache-dir "transformers>=4.51.0" "accelerate>=0.34.0"
COPY entrypoint.sh /entrypoint.sh
RUN chmod +x /entrypoint.sh
EXPOSE 7860
ENTRYPOINT ["/entrypoint.sh"]