[Bart] Move CLS rep extraction from EOS tokens to head classes #624
Conversation
This file contains 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
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Outdated suggestions cannot be applied.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.
Fixes #494 and fixes #563.
BartAdapterModelcurrently tries to extract CLS token representations from EOS tokens independent from the used heads (if at all). This causes weird issues for models not using CLS token based heads.This PR moves CLS token rep extraction into the specific head classes to only compute it when needed. Two (re-usable) new args to
forward_head()are introduced:get_cls_from_eos_tokens: to indicate a model uses CLS reps from EOS tokenseos_mask: token mask to extract CLS reps (required when passingget_cls_from_eos_tokens=True)