Testing and continuous integration

Every push and pull request to main is automatically tested by GitHub Actions (CI workflow): a lint pass for critical errors plus the pytest suite on Python 3.10–3.13, covering unit tests (metrics, backbones, datasets, API) and a full CPU end-to-end smoke run of train_pa → train_dpd → run_dpd → plot (including W16A16 quantization-aware training). A scheduled Weekly workflow additionally trains every supported backbone and every built-in dataset on Linux and macOS. The badges at the top of this page reflect the latest results.

To run the tests locally:

pip install -e ".[dev]"
pytest -m "not extended"   # fast suite (~1 min on CPU)
pytest                     # everything, incl. all backbones / all datasets

Documentation site

This site is built with MkDocs and Material for MkDocs from the docs/ directory and deployed to GitHub Pages by the Docs workflow on every push to main. The pages are single-sourced: they include sections of README.md, datasets/README.md, examples/README.md and benchmark/benchmark_report.md between --8<-- markers, so editing those files updates the site. To preview it locally:

pip install -r docs/requirements.txt
mkdocs serve

mkdocs build --strict is what CI runs: a broken link or a missing include fails the build.