Adam Li

Merging. Thanks for the in-depth review and fixes @PSSF23 !

Notes: 1. fused-type util functions to allow vector[intp_t] and intp_t[:] is useful and should be merged upstream 2. the performance penalty of MORF most likely is in space and in...

Yeah I think because we do imports in `__init__.py` I would take a look at the strategy, or designs sklearn uses to soft-import optional dependencies. E.g. polars, or joblib

Just to clarify @ryanhausen rather than warning during import, it should only warn the user when the function is used

> I am resurrecting this, but for whoever would like to work on trees, I think the approximate splitter contribution should be assigned high priority. This should not be that...

Happy new years! Thank you for this detailed response. Looking forward to additional responses from the training team! Some follow-up questions after reading thru the [benchmarks you linked](https://developer.nvidia.com/blog/accelerating-random-forests-up-to-45x-using-cuml/#benchmarks). No rush...

Hi just wanted to follow up with your team on the thread regarding "training time speedups" now that the holidays are over. Happy to clarify if any questions. Thanks!

Hi @wphicks just following up. Thanks!

I see! That's exciting. I always did find the `parametrize_with_checks` just a blackbox, so I think having some more fine-grained control over what's "checked" is useful.

Few questions for someone who doesn't understand the context. Would the idea be for any 3rd party inheriting from `BaseEstimator` to have to define their own `__validate_data__` function, else it...