mirror/github-mirror.sh publishes main and the tags on it to a read-only
GitHub mirror, as one extra commit that adds a notice under the README
title, so every forge commit is unchanged there and each sync is a
fast-forward. A systemd user service and hourly timer run it from a host
that can push to GitHub; --settings turns off issues, pull requests and
the rest there.
A GitHub Actions workflow builds the image on the mirror and pushes it to
ghcr.io, freeing runner disk first since the CUDA build of torch is
several GB. docker-compose.yml names the image after IMAGE_PREFIX and
IMAGE_TAG, so setting them pulls the published image instead of
building, and the README says how.
python:3.11-slim base relies on pip torch wheels bundling CUDA runtime libs,
with the host NVIDIA driver passed through via the nvidia container runtime
(compose GPU reservation); TORCH_HOME and XDG_CACHE_HOME point into the
mounted /data volume so Demucs/wav2vec2 weights survive container
recreation.
Docker was not installed in the implementation sandbox, so the image was
not built or run and remains untested. PyYAML was also unavailable, so
docker-compose.yml was not parse-checked programmatically — only hand
reviewed.