Thomas J. Fan

If we do not support bytes, we should error and consider this PR is a bug fix. If we do support bytes in either form: `np.array( BYTES, dtype="S")` or `np.array(BYTES,...

@cozek I would leave this PR as it for now. We need to decide if we should support classification targets encoded as bytes I'll bring it up during the next...

@glemaitre This PR has changed a bit since you approved it. Now the PR deprecates the bytes format as the target instead of raising an error. Are you okay with...

For `PoissonRegressor`, is it enough to define a `decision_function` that outputs the raw scale? Then we can use `response_method="decision_function"` with `partial_dependence` to get the desired plots.

With the recent inclusion of executable examples in the browser using JupyterLite in https://github.com/scikit-learn/scikit-learn/pull/25887, I prefer for meson to work with Pyodide before switching over. Currently there is work to...

Progress update as of a few weeks ago: I got `cirun.io` working on my fork, but I ran into an issue with it stalling with PyTorch. I suspect it was...

The latest work is at https://github.com/thomasjpfan/scikit-learn/tree/cirun_gcp (GCP) I'll take another took at it this weekend. If I do not have a PR by Monday, then let's schedule a call so...

After a few more iterations, it works now! https://github.com/thomasjpfan/scikit-learn/actions/runs/7820229387/job/21334434689 Next step is to to wrap it up into our workflow for triggering cirun.

It also works with T4s: https://github.com/thomasjpfan/scikit-learn/actions/runs/7820372944/job/21334925253 it looks to spin up faster, because there are more T4s available.

I'm +1 on at least having a weekly run. The hard part comes from PRs. I proposed using a "run-gpu-ci" GitHub label to trigger the CI. Here are the details:...