[MAINT] Update pinned sklearn and numpy

[chiuhoward]

numpy==2.2 and sklearn==1.6.1

[chiuhoward]

@arokem I think we’re up against a pretty big hurdle but that has been well-documented. If I’m understanding this correctly we’re just waiting for the next stable release of xgboost? (or use sklearn<1.6.1 which kind of defeats the purpose of this branch)

scikit-learn/scikit-learn#30542

The other test failure is due to neurocombat_sklearn where I believe we need to change sparse to sparse_output for OneHotEncoder.

https://github.com/Warvito/neurocombat_sklearn/blob/25306f0a2f088764b40a376ba276a2126ff820b4/neurocombat_sklearn/neurocombat_sklearn.py#L191

[arokem]

Yes - I think that we need to wait for the XGBoost release to update all the way, but I think that we can at least advance to 1.5.2 for the time being, so we can keep moving here. I don't think that defeats the purpose altogether, as will ease the compatibility with pyAFQ by allowing use of numpy 2.0 versions. Once we have that under control, then yes, we should move on to advance #15, where we vendorize neurocombat_sklearn and fix it here.

…

On Fri, Jan 31, 2025 at 11:08 PM Howard Chiu ***@***.***> wrote: @arokem <https://github.com/arokem> I think we’re up against a pretty big hurdle but that has been well-documented. If I’m understanding this correctly we’re just waiting for the next stable release of xgboost? (or use sklearn<1.6.1 which kind of defeats the purpose of this branch) scikit-learn/scikit-learn#30542 <scikit-learn/scikit-learn#30542> The other test failure is due to neurocombat_sklearn where I believe we need to change sparse to sparse_output for OneHotEncoder. https://github.com/Warvito/neurocombat_sklearn/blob/25306f0a2f088764b40a376ba276a2126ff820b4/neurocombat_sklearn/neurocombat_sklearn.py#L191 — Reply to this email directly, view it on GitHub <#20 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAA46NXPTXZNZPIWMK6KX332NRXH5AVCNFSM6AAAAABWITJZASVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMMRYHAZDCMJTGA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[arokem]

This is now green! 🙌 Any reason not to merge?

[arokem]

One remaining challenge is that pypi will not let us upload our software with a GitHub dependency: https://github.com/tractometry/AFQ-Insight/actions/runs/13125078725/job/36619634880#step:6:174, so I think that we still need to vendorize neurocombat_sklearn into our software, which I will attempt on a follow-up PR.