What's Changed
- Full PhiO pre/post-processing support for text, vision, and audio
- RegEx loading from tokenizer.json
- Custom RegEx processing for tokenizer text splitting
- Support for default inputs in Python decoder API, fixing #852 and related issues
- Handling for models with mismatched added_tokens and added_tokens_decoder
- Support for left and right strip for added tokens
Full changelog: v0.13.0...v0.14.0