Skip to content

Conversation

frgfm
Copy link
Contributor

@frgfm frgfm commented Aug 20, 2022

As discussed in #6439, this PR adds support for the poly loss. A few things are left to decide / do:

  • decide whether a module API should also be added (easy to add in this PR)
  • specify the depths of the unittests (I've checked the one of the focal loss, it's quite extensive. how much should we do? checking trivial vaues?)

Closes #6439

cc @datumbox

Copy link
Contributor

@oke-aditya oke-aditya left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sorry for early poke but a few thoughts @frgfm

@datumbox
Copy link
Contributor

@frgfm Just wanted to check in and ensure you are not blocked or need any assistance. Let me know if you do. :)

@frgfm
Copy link
Contributor Author

frgfm commented Sep 16, 2022

Hello there :)

My bad, I just caught up with the PR! I left it as a draft because the missing part is the unit test: I saw that the ones for the focal loss are quite extensive. Do I need to be that thorough?

@datumbox
Copy link
Contributor

@frgfm Ah yes the unit-tests. I would recommend following the pattern that we have on existing tests and have a similar coverage. Hopefully following the existing infra will make your work easier. The reason we really need them is because previously we spotted a few nasty bugs on some of the losses. We are only couple of weeks prior cutting the branch for the upcoming release so if we want to include this feature we need to have elevated degree of confidence.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Add support for PolyLoss in torchvision

4 participants