Multi-arch (amd64 + arm64) Docker images for running vLLM with CPU backend.
Built automatically from upstream tags and commits.

• nightly → current commit of main
• <version> (e.g. 0.10.2) → official vLLM release tags
• <commit-hash> → alternative tag for reproducibility
• amd64-*, arm64-* → per-arch builds
• latest → alias to latest release

Run an API server:

docker run --rm -p 8000:8000 \
  -e HUGGING_FACE_HUB_TOKEN=\$HF_TOKEN \
  gabrielbico/vllm-cpu:nightly \
  --model google/gemma-3-270m

• linux/amd64 (built on AVX2 hosts)
• linux/arm64 (Apple Silicon, ARM servers)

These images are CPU only. For GPU builds, see the official vLLM docs.