Fix issue for alignment of pre-tokenized input with BPE, see #150 #154
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
|
Update: tested and good with some BERT models, some Roberta models, CamemBERT (although it is Roberta-based, camembert uses a SentencePiece tokenizer using Note: fnet-base and luke-base are not supported by AutoModel apparently |
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Outdated suggestions cannot be applied.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.
Fix an issue with added extra tokens for ensuring the alignment of pre-tokenized input with BPE, see #150
We cover here added tokens introduced by the BPE fallback mechanism for out of vocabulary input character, e.g. "É" resulting in 'Ã', 'ī', which have then confusing offsets when the input is pre-tokenized.
We distinguish transformer tokenizers using BPE/sentencepiece from the other, which is done based on the tokenizer class name. When BPE is used, we examine offsets and tokens in order to filter out added tokens due to the BPE OOV fallback mechanism.
Tested with BERT based models and Roberta based models, more test to come...