Nicolas Hug

I don't think there's anything to deprecate here: there's no feature, just a silent (minor) bug. We could raise a temporary warning instead of an error, but for such bugfixes...

@thomasjpfan what made you change your mind from error to warning? The offending test [sparse_svm](https://github.com/scikit-learn/scikit-learn/blob/1045d16ec13b1cab7878e7555538573d1884aad3/sklearn/svm/tests/test_sparse.py#L286) is clearly wrong as it passes `clf = svm.OneClassSVM(gamma=1, kernel=kernel)` for all kernels.

Hi @Isalia20 , thanks for the PR. This is a *massive* one :) Can you share more about why this is needed? I shared my thought and questions on that...

Hi @Isalia20 . I just want to be fully transparent to manage expectations: we won't be able to make this happen for 2.10. Reviewing this PR is a massive amount...

@AntoineSimoulin and I discussed this a bit offline, my current thoughts are that: - we shouldn't expect users to pass masks to `labels_getter` for them to be sanitized, that would...

Hi @zhumakhan can you share more on what exactly doesn't work? What's the error message?

Thanks for opening this issue @Brokttv I am not sure such a tool would be in scope for torchvision, considering the infinitely vast input (and output) space of such a...

Hi @sandeepgupta12 , thank you for the proposal. I see there are similar discussion on the torch repo about this https://github.com/pytorch/pytorch/issues/141235#issuecomment-3160480841 What we can do in torchvision will depend on...

Thanks for the report @Alek96 . I don't know/remember why nearest-exact wasn't exposed for masks in v2, but we should probably enable it.

For context on nearest vs nearest exact: https://github.com/pytorch/vision/issues/6645 > Mode mode='nearest-exact' matches Scikit-Image and PIL nearest neighbours interpolation algorithms and fixes known issues with mode='nearest'. This mode is introduced to...