I am noticing that some of the tests are failing due to scikit-learn 1.7.0, while scikit-learn is pinned to versions 1.5.0 and 1.6.0 in tests.
I agree this is wise - scikit-learn has been changing extension contracts recently quite often, leading to breakage in extender packages (such as feature-engine) with the minor releases.
My thoughts were, when I saw this, perhaps it would be useful to have a test matrix here? Perhaps:
- 1st test matrix with most recent
sklearn, and matrix of operating systems (win, mac, unix) and python versions
- 2nd test matrix with different
sklearn minor releases 1.5.0, 1.6.0, 1.7.0, etc, one operating system, multiple python versions
What do the core devs think?
I am unfortunately not familiar with circle CI, but I would know how to add this via GHA.
I am noticing that some of the tests are failing due to
scikit-learn 1.7.0, whilescikit-learnis pinned to versions 1.5.0 and 1.6.0 in tests.I agree this is wise -
scikit-learnhas been changing extension contracts recently quite often, leading to breakage in extender packages (such asfeature-engine) with the minor releases.My thoughts were, when I saw this, perhaps it would be useful to have a test matrix here? Perhaps:
sklearn, and matrix of operating systems (win, mac, unix) and python versionssklearnminor releases 1.5.0, 1.6.0, 1.7.0, etc, one operating system, multiple python versionsWhat do the core devs think?
I am unfortunately not familiar with circle CI, but I would know how to add this via GHA.