fix: exclude last position from loss_mask after target shift#341
Open
AchuthReddy-16 wants to merge 1 commit into
Open
fix: exclude last position from loss_mask after target shift#341AchuthReddy-16 wants to merge 1 commit into
AchuthReddy-16 wants to merge 1 commit into
Conversation
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 #317
target and input_ids are shifted left by one position in dataprepare()
via padding(..., left=False), which puts a zero-padding value in the
last position instead of a real next-token value. loss_mask wasn't
adjusted to match, so training was computing loss on that final,
invalid position.
This adds one line to exclude the last position from the mask after
the shift, as originally suggested in the issue.
Verified with a small repro (numpy, mirrors the actual padding() logic):
before fix
target: [202 203 204 205 0]
loss_mask: [1 1 1 1 1] <- last position still marked for training
after fix
target: [202 203 204 205 0]
loss_mask: [1 1 1 1 0] <- correctly excluded