Peter Andreas Entschev

It's important to add that this is the behavior today, unless the user specify a limit. In other words, we would need to change the default behavior to match that...

This should be fixed now, closing.

Oops, my mistake. Thanks for catching this @jakirkham .

Thanks @otavioon for the report. You're right, they're currently not supported, my guess is that this was copied from `dask-worker` and never modified. Are those options important for you?

I'm sorry @otavioon , I thought I had replied to this issue but it probably got lost somehow. Would you be interested in contributing those changes to Dask-CUDA, it would...

I think this is a good starter issue. If any of you would like to work on it, contributions are much appreciated! :)

Have you had a chance to test https://github.com/dask/distributed/pull/3670 @VibhuJawa ? Is this still an issue?

The build will fail now, since it depends on NumPy and CuPy fixes that weren't yet merged. However, I can confirm that all existing tests pass on my setup after...

FWIW, I have no strong opinions there. Sorry if this question has already been answered before or sounds a bit stupid, but wouldn't it then make sense to have Dask-GLM...

As far as I recall, this PR should mostly work (if not entirely). It's not necessarily dependant upon #66, even though a little modification would be required once it's merged....