ablaom
commented
5 months ago I'm happy to approve this PR after appropriate testing (see below). However, it's is essentially a patch on top of a hack, for which I am personally to blame. That said, I think there is a fair chance that the overloading of in for MLJType (which is at least 3 years old, according to the blame) was rendered redundant when I refactored the learning networks code. The only way to find out for sure, would be to remove it and locally run
- MLJBase tests
- MLJTuning tests
- MLJ#dev tests, with
MLJ_TEST_INTEGRATION="true".
Is this something you would consider doing? Even if we accept the current PR, we probably want to run those tests anyway, so it's the same work really.
This is needed to fix some failing test in another package. See https://github.com/JuliaAI/FeatureSelection.jl/actions/runs/9059544029/job/24887362493?pr=5#step:6:523