EmilyXinyi

I am also trying to reproduce the problem. I am on Mac with Intel i5 cores

Colab: https://colab.research.google.com/drive/15PM-xqXoSME7h5N9RcPI2OQv3qK7Fpvi?usp=sharing

> The upcomping scipy 1.14.0 release will ship a array API support for this function via a generic fallback. > > https://github.com/scipy/scipy/blob/3aa715425291727cfd63d3929802d242681318ba/scipy/special/_support_alternative_backends.py#L76-L81 > > Since it is not released yet,...

I have made the suggested changes to `azure-pipelines.yml` and `.github/workflows/cuda-gpu-ci.yml`, and here is the colab link for cuda and nightly build of scipy: https://colab.research.google.com/drive/1Dw6ABHw-2wyjlzUQKb4VDJw_51mGZMLn?usp=sharing Seems like the tests are passing...

> @EmilyXinyi the `main` has evolved concurrently and there is now a bunch of conflicts. Would you mind updating the PR to resolve those. Please let me know if you...

> Should we instead check when enabling array API support and print a warning to the user? Should this change be made this PR? (If yes, could anyone help me...

Just a note: please ignore all the recent commits and force pushes. I think I did a merge instead of rebase and a bunch of files I didn't touch changed.

> We might need to add a `--upgrade` or `--force-reinstall` flag to that command. Or both flags. > > Then check the output of the cell with `sklearn.show_versions()` to make...

Hello @ogrisel and @betatim I finally fixed this issue and ready for another review. I added some tests to cover the `UserWarning` case (otherwise the pipeline fails) and the colab...

I closed this by accident while fixing merge conflicts ;-; if a maintainer could reopen this PR it would be great, thank you!