Skip to content

contributing.md

Latest commit

 

History

History
81 lines (58 loc) · 3.89 KB

contributing.md

File metadata and controls

81 lines (58 loc) · 3.89 KB

Contributing

Contributions are welcome! Here are some ways to help:

  • Add a new backbone — The Braindecode library always welcomes new backbone architectures! Follow the model requirements and submit a PR to braindecode's GitHub with your backbone architecture. Then add a factory function in default_configs/backbones.py that loads your new model.
  • Add a new dataset — Push a pre-windowed BaseConcatDataset to HuggingFace Hub and add a factory function in default_configs/datasets.py
  • Add a new fine-tuning strategy — Implement a new class in finetuning.py inheriting from BaseModel with apply() and get_callbacks() methods
  • Report bugs or suggest features — Open an issue on GitHub

Development setup

uv pip install -e ".[dev]"
pytest  # make sure everything passes

Running Tests

pytest                                # all tests
pytest tests/test_default_configs.py  # config validation
pytest tests/test_backbones.py        # model build & forward
pytest tests/test_normalization.py    # normalization correctness
pytest -k "test_name"                 # single test

Releasing a new version

Releases follow semantic versioning. Replace X.Y.Z below with the new version (e.g. 0.3.0) and YYYY-MM-DD with today's date.

1. Open a release PR from a dedicated branch

git checkout main && git pull
git checkout -b release/X.Y.Z

2. Bump metadata

  • pyproject.toml — set version = "X.Y.Z".
  • CITATION.cff — set version: X.Y.Z and date-released: "YYYY-MM-DD" (these are the software version and release date; do not touch cff-version, which is the CFF schema version).
  • CHANGELOG.md:
    • Rename the existing ## [Unreleased] header to ## [X.Y.Z] - YYYY-MM-DD.
    • Insert a new empty ## [Unreleased] section above it (required — every PR must update the changelog, so there must always be an Unreleased section).
    • At the bottom of the file, update the [Unreleased] link to compare/vX.Y.Z...HEAD and add a new [X.Y.Z]: https://github.com/braindecode/OpenEEGBench/releases/tag/vX.Y.Z line.

3. Refresh the lock file

uv lock

4. Commit, push, open the PR, wait for CI

git add -u && git commit -m "Version X.Y.Z"
git push -u origin release/X.Y.Z
gh pr create --base main --title "X.Y.Z" --body "Release X.Y.Z"

Wait for all checks to pass, then squash-merge into main.

5. Create the GitHub release

On the releases page:

  • Tag: vX.Y.Z (with the v prefix — required by the CHANGELOG.md links, by previous tags, and by the Zenodo archive URL).
  • Target: the squash-merge commit on main.
  • Title: vX.Y.Z.
  • Description: paste the ## [X.Y.Z] section from CHANGELOG.md, followed by **Full Changelog**: https://github.com/braindecode/OpenEEGBench/compare/vPREV...vX.Y.Z.
  • Click Publish release (not "Save draft" — the PyPI and Zenodo triggers only fire on published).

Clicking Publish automatically:

  • runs publish.yml → uploads the sdist + wheels to PyPI and attaches them to the GitHub release;
  • triggers the Zenodo webhook → archives the tagged snapshot and mints a new DOI under the concept DOI.

6. After publication

  • Verify the package appears on PyPI.
  • Verify the release appears on Zenodo with the expected metadata (authors, affiliations, ORCIDs).
  • On the first release only: replace the TODO placeholders in README.md (DOI badge and doi field in the BibTeX block) with the concept DOI from Zenodo, and open a follow-up PR.