Provides an unofficial Docker image with samapi. Images are generated monthly.

Please cite the original repository if you use this docker image.

docker run \
    --gpus all \
    -p 8000:8000 \
    -v ~/.samapi:/.samapi \
    ghcr.io/siemdejong/samapi-docker

• select gpus with --gpus ...
• map ports with -p ...:8000
• mount a local directory (e.g. ~/.samapi) to /.samapi to cache model weights.
• pass --workers=... to change the default number of workers (2).

Images have are tagged by cuda version (*-cu118 or *-cu128). latest always refers to *-cu118 to ensure compatibility.