FROM roacoh/vllm-cpu-env ARG VLLM_CPU_KVCACHE_SPACE ARG VLLM_CPU_OMP_THREADS_BIND ENV VLLM_CPU_KVCACHE_SPACE=${VLLM_CPU_KVCACHE_SPACE} ENV VLLM_CPU_OMP_THREADS_BIND=${VLLM_CPU_OMP_THREADS_BIND} RUN --mount=type=secret,id=HF_TOKEN,mode=0444,required=true \ export HF_TOKEN=$(cat /run/secrets/HF_TOKEN) EXPOSE 8000 ENTRYPOINT [ "vllm", "serve", \ "--model", "google/embeddinggemma-300m", \ "--dtype", "bfloat16", \ "--host", "0.0.0.0", \ "--port", "8000" ]