Vasilis Vryniotis

Potentially related or duplicate to #5419

@lezwon thanks a lot! It's worth coordinating with @vfdev-5 because he is literally landing changes at the same API. Victor, how do you propose to do this? Would you prefer...

@ain-soph I think it would be awesome if you implement it. Just checking again with @vfdev-5 and @pmeier that they don't have any concern.

@ain-soph that's exactly what we need to do. We should have already all the ops excluding perhaps the `*_Only_BBoxes`. Here are examples of the codebase for these implementations: https://github.com/pytorch/vision/blob/becaba0e0546f434be64de4e9603ab66e4b160b2/torchvision/prototype/transforms/functional/_geometry.py#L259-L268

@ain-soph So far the kernels process each input independently. Aka they don't receive together `img` and `bounding_box` as inputs, but instead usually contain one of them along with configuration that...

@ain-soph I was hoping you could confirm that Victor's reply unblocked you and you are able to continue the feature. Please let me know if there are more outstanding questions....

I have a use-case that requires LabelSmoothing. Unfortunately `CrossEntropyLoss` does not support it in PyTorch (https://github.com/pytorch/pytorch/issues/7455). This is a highly requested feature but unfortunately it's been blocked for more than...

@yassineAlouini Wow what a coincidence! Today I was working on something related. :) At #5444, I have an [experimental private function](https://github.com/pytorch/vision/pull/5444/files#diff-5288f1a12d4cbc77afac42d57b75796da3c8734c0fa4cda80e5c07ac5d16212dR513) that makes it possible to switch between losses. There...

@oke-aditya @yassineAlouini It would be awesome if you could help on the development and review of these 2 losses. For now will put them flat on the ops package similar...

@yassineAlouini What Aditya said ^. :) No need for a separate ticket, we got plenty that mention it already. When you bring the PR, I'll tag it accordingly. Just make...