Conversation
rufinv
approved these changes
Jan 8, 2026
rufinv
approved these changes
Jan 8, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adding the changes from the attention paper, where they are relevant !
the general idea is I :
removed the fused loss (it now counts as dcy)
we still log all the losses individually (by encoded and decoded modalities)
separated the cycle loss from the broadcast loss function (without re-encoding anything, the broadcast loss returns the elements required for cycling)
added the paper's attention version + a helper for the user to switch attention to trained after training a RandomSelection gw
added tests for stability.
we decided against including the branching between different selection mechanisms inside of the broadcast loss, which was only useful for the case where we were doing augmentation on MM-IMDB