Fix BERTScore OverflowError when the tokenizer has no model_max_length - #790
Open
tonycoder-hub wants to merge 1 commit into
Open
Conversation
Tokenizers whose config omits `model_max_length` report transformers' VERY_LARGE_INTEGER sentinel (~1e30) instead. bert_score forwards that value to `tokenizer.encode(max_length=...)`, which overflows the Rust tokenizers backend used by transformers>=5, e.g. for microsoft/deberta-xlarge-mnli. Cap the sentinel to 512 before any tokenization happens, and add a `max_length` argument so a different truncation length can be requested. Fixes huggingface#739 Co-authored-by: Tony Coder <407243179@qq.com>
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 #739
Problem
bertscore.compute(..., model_type="microsoft/deberta-xlarge-mnli")crashes withOverflowError: int too big to convertundertransformers>=5. Tokenizers that omitmodel_max_lengthreport a huge sentinel, andbert_scoreforwards it intotokenizer.encode(max_length=...).Fix
Cap the sentinel to 512 (BERT-family default / DeBERTa
max_position_embeddings) before any tokenization, including theidf=Truepath. Expose an optionalmax_lengthoverride. Models that already declare a realmodel_max_lengthare untouched.#756 proposed a similar cap but applied it after
BERTScorerconstruction, soidf=Truestill overflowed.Tests
test_bertscore_caps_undefined_model_max_lengthandtest_bertscore_max_length_overrides_model_max_lengthintests/test_metric_common.py(idf=True).python -m pytest tests/test_metric_common.py -k bertscore -q: 3 passed.make qualityclean on the changed files.