fix: byte fallback penalty in SentencePiece Viterbi tokenizer#4
Merged
Conversation
Byte fallback tokens (<0xNN>) were competing with multi-character vocab tokens in the Viterbi DP using their actual vocabulary scores. When byte token scores happened to be higher than vocab token scores, the Viterbi algorithm preferred 43 byte-level tokens over 7 word-level tokens. Fix: assign byte fallback tokens a fixed score of -1e6 instead of their vocabulary score, ensuring they are only used as a last resort when no vocab token covers a position. This matches llama.cpp behavior.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Byte tokens used fixed -1e6 penalty instead of vocab scores. Fixes Mistral producing 43 byte tokens instead of 7 word tokens.