Enable Tokenizers with Byte Tokens #1153
Merged
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Outdated suggestions cannot be applied.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 #1038
Problem
Qwen directly uses
bytes
as tokens for "incomplete unicode". We don't handletype(token) == byte
infsm.py
.For "incomplete unicode" handling with
llama
andgpt2
style BPE tokenizers,fsm.regex.reduced_vocabulary()
looks for a string prefix indicating it's incomplete unicode (e.g.\ufffd
), converts the token string to bytes, and strips the prefix "incomplete unicode indicator" bytes to get thetoken_bytes
.Change
For this PR's "incomplete unicode" handing for "Qwen-style" BPE, the token is already
bytes
, so if isinstance(token_str, bytes): token_bytes = token`.