Automated Testing Across Multiple scikit-learn Versions

[Gnpd]

Set up automated tests to ensure compatibility of OpenModels with different versions of scikit-learn. Implement a CI workflow (e.g., using GitHub Actions or similar) that runs the test suite against a matrix of scikit-learn versions. This will help detect breaking changes early and maintain support for a range of scikit-learn releases.

For local we can use tox or manually:

python -m venv .venv-sklearn-1-5
.venv-sklearn-1-5\Scripts\activate
pip install scikit-learn==1.5.2
pip install -e .
pytest

Ensure that [tool.poetry.dependencies] is set to the version you want to test scikit-learn = "1.5.2"

[Gnpd]

Resolved in PR #52.

I implemented a dedicated GitHub Actions workflow (.github/workflows/sklearn-compat.yml) that runs the full test suite against sklearn 1.6.1, 1.7.2, and 1.8.0 independently of the main CI pipeline.

How it works:

• A matrix of 3 jobs (one per sklearn version) runs on a fixed Python 3.12 environment
• Dependencies are installed via poetry install, then sklearn is overridden with pip install "scikit-learn==X.Y.Z" --force-reinstall — no new tooling required
• fail-fast: false ensures a failure on one version does not cancel the others
• Triggers on push to main (when source or tests change), weekly on Mondays, and manually via workflow_dispatch

All three versions passed. The README now includes a compatibility matrix so users know which versions are covered by automated tests.

Testing locally (no tox needed):

poetry install
poetry run pip install "scikit-learn==1.6.1" --force-reinstall
poetry run pytest

Repeat for 1.7.2 and 1.8.0. Restore with poetry install when done.