Release?

[jrbourbeau]

The 2021.11.0 release of dask recently broke some functionality in dask-ml which has since been fixed (xref #878). It's probably worth pushing out a dask-ml release which includes that patch. What do others think? FWIW I'm happy to handle releasing if @TomAugspurger is busy with other things (Tom, does dask-ml use the usual dask/dask release process, or are there additional / different steps?)

There's also #862 which has been active recently and looks like it's close to being completed. Should we wait for that before releasing?

[TomAugspurger]

I think #862 is close. I think it's worth waiting a day or two to get that in.

My last round of CI changes updated the release stuff. Aside from a changelog, a Dask-ML release should just require making a release on GitHub. Then the release action takes care of building a wheel and sdist and uploading it to PyPI (might be worth trying that for dask/dask or dask/distributed, if you want to avoid having to tag / build wheels locally).

[TomAugspurger]

dask-ml 2021.11.16 is released to PyPI with https://github.com/dask/dask-ml/releases/tag/v2021.11.16. Thanks James!