Starstruk Logo MarkStarstruk

v0.15.3

datalab-to/suryav0.15.3Aug 9, 2025by VikParuchuri

AI Summary

Users can now fine-tune the model with a dedicated script, and the OCR model has been updated for slightly better accuracy.

Key Highlights

  • Added finetuning script for easy model training
  • Bumped OCR model for better accuracy
  • Added progress bar for model downloads
  • Fixed tokenizer for script tokens

New Features

  • Fine-tuning support
  • Download progress bar

Full Release Notes

## Finetuning support

- Add support for finetuning, and a script you can run to finetune the model easily

## OCR improvements

- Bump the OCR model to a newer version that is slightly more accurate

## Misc

- Add progress bar for model downloads


## What's Changed
* feat: download progress bar for each file by @zanussbaum in https://github.com/datalab-to/surya/pull/421
* Fix tokenizer to correctly tokenize script tokens by @starikovplusplus in https://github.com/datalab-to/surya/pull/423
* [WIP]: Finetuning Script by @tarun-menta in https://github.com/datalab-to/surya/pull/422
* Dev by @VikParuchuri in https://github.com/datalab-to/surya/pull/424

## New Contributors
* @starikovplusplus made their first contribution in https://github.com/datalab-to/surya/pull/423

**Full Changelog**: https://github.com/datalab-to/surya/compare/v0.15.2...v0.15.3