jmakov

Can't this be simply parallelized with ray.io?

Interesting, thought one would just refactor and add local ray cluster backend: https://github.com/autogluon/autogluon-cloud.

My issue might be related: https://github.com/ray-project/ray/issues/39565

Or https://github.com/narwhals-dev/narwhals (extensible compatibility layer between Polars, pandas, Modin, and cuDF). Guess the winner would be https://github.com/Eventual-Inc/Daft - Polars on Ray.

Guess one could ask the guys at https://github.com/TracecatHQ/functime how they managed to do it.

If I set `preprocess=True` and omit the `pca*` args, then I see again all the models being compared. Also seems that including categorical data was problematic - works as expected...