Fixing a pathological case for slow tokenizers #14981
Merged
+14
−2
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
Narsil
commented
Dec 29, 2021
stevhliu
pushed a commit
to stevhliu/transformers
that referenced
this pull request
Jan 6, 2022
* Fixing a pathological case for slow tokenizers * Update src/transformers/tokenization_utils.py
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Outdated suggestions cannot be applied.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.
What does this PR do?
Fixes issue found here for slow tokenizers:
huggingface/tokenizers#848
When using arbitrary tokens, a bug could occur where we would seed the new character, even when the lookahead attempted to skip over that part of the text. Thus we could have an extra match that didn't fit.
Other bug was during the lookahead we could undermatch (since we were iterating on new characters before checking termination).
The test added covers both cases.
Fixes # (issue)
Before submitting
Pull Request section?
to it if that's the case.
documentation guidelines, and
here are tips on formatting docstrings.
Who can review?
Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.