v0.4.3
allenai/olmocrv0.4.3Oct 31, 2025by github-actions[bot]
AI Summary
v0.4.3 adds support for finetuning with new configurations, improves the priority scheduler, and introduces benchmarking enhancements including support for Cirrascale and DeepInfra providers.
Key Highlights
- Finetuning support added
- New benchmarking providers (Cirrascale, DeepInfra)
- Priority scheduler fix
- Benchmarking scripts
New Features
- Finetuning capabilities
- Benchmarking infrastructure
- Priority scheduler
Full Release Notes
## What's new ## Commits 3990b94 Bump version to v0.4.3 for release 419b576 Version bump a39ca53 Merge branch 'jakep/priority_scheduler' e95aaed Fixing issue which likely corrects #371 e1e3d5b README for finetuning ec1bf24 Removing $ inline math from dolma viewer script, it's not reliable and our models don't output that format anymore 3d2c977 Merge branch 'jakep/lorafix' 41bee1e Update README.md 4768db6 Fixes to train.py 91962d6 Example finetuning config c62e221 Adding init file in data 4ee2005 Adding a flag to run_benchmark script 579e52a Adding Cirrascale as inference provider 247d51d Updating deepinfra pricing 10ab6a6 Adding suppor for smaller error bars in benchmarking 15496b9 Infrapartner benchmarking script 2d2c2c9 Adjusted table 7d551ec Cleanup bench results table 7c5a385 Copy over results table to main readme 2c90e6a More readme updated with paper benchmark scores 53be91a Add citations and arxiv paper links ab108a5 Lint fixes 210389f Merge branch 'main' of https://github.com/allenai/olmocr 5083967 Cleaning up repo to move all unit tests to a consistent place