add feature_names_in_

[tvdboom]

Closes #5677

Adds the feature_names_in_ attributes to all estimators. Code copied from sklearn's BaseEstimator.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[csadorf]

/ok to test

[tvdboom]

@csadorf can you run the build?

[csadorf]

/ok to test

[csadorf]

Tests are currently failing because of a missing import. I would appreciate if you could fix that before I do a more thorough review.

[csadorf]

@tvdboom Let me know if there is any way I can assist you with this PR!

[tvdboom]

I haven't found the time to work on this yet. I'll do it for sure next week. thanks

[csadorf]

I haven't found the time to work on this yet. I'll do it for sure next week. thanks

Sure, no rush at all! Just let me know when you want to move forward with this.

[tvdboom]

@csadorf It should be ready now. I imported numpy over cupy. I believe for this specific array that's probably faster but let me know if you think cupy would be the better choice here.

[csadorf]

/ok to test

[csadorf]

/ok to test

[csadorf]

Looks good overall, thanks a lot!

I ran some tests and also noticed that the warnings module is used in the skl_dependencies.py module, but is not actually imported. That would of course need to be fixed, but I'm also a bit concerned, because it means that those branches are not actually tested. I would appreciate if you could have a look at how much effort it would take to test those cases.

[csadorf]

/ok to test

[csadorf]

@tvdboom Please do not merge the remote tracking branch unless necessary. It stresses our CI resources. Also, please install the pre-commit hooks to ensure that style checks will pass. Thanks!

[csadorf]

/ok to test

[csadorf]

@tvdboom Looking at the test outputs, it seems like it's not working as expected. Do you have any idea what might be going on? Do you need assistance in setting up a local test environment for debugging?

[tvdboom]

Unfortunately, I have a new pc with an amd gpu and windows os so I can't install cuml locally. So what I did was to run the tests importing a model from sklearn instead of cuml and (naively) hope that this would also work with the same model from cuml. Apparently it does not, so not sure how to debug.

[csadorf]

Unfortunately, I have a new pc with an amd gpu and windows os so I can't install cuml locally. So what I did was to run the tests importing a model from sklearn instead of cuml and (naively) hope that this would also work with the same model from cuml. Apparently it does not, so not sure how to debug.

That's unfortunate. Maybe you can use Google Colab?

[tvdboom]

That's unfortunate. Maybe you can use Google Colab?

Any idea how to install cuml from a custom branch in colab? I tried this:

Also adding --extra-index-url=https://pypi.nvidia.com gives the same error.