docling-project / docling-project/docling
Address transformers "false-alarm" warning in hybrid chunker
Open
chunker
enhancement
triage/close-stale
- Dominant language
- Python
- Stars
- 66.4k
- Forks
- 4.8k
- Avg merge
- 2d 21h
- Merged PRs (30d)
- 84
Description
### Requested feature
Address the "false-alarm" message from the transformers library emitted during hybrid chunking & discussed [here](https://github.com/DS4SD/docling-core/issues/119).
This could be done by catching the warning using [this](https://docs.python.org/3/library/warnings.html#testing-warnings).
HuggingFace is using a variant where they emit an additional warning, calling to ignore the previous warning ([ref](https://github.com/search?q=repo%3Ahuggingface%2Ftransformers+%22token+indices+sequence+length+is+longer%22&type=code)).
Contributor guide
Assessment
This issue has not been assessed yet.