What's Changed
- Revert "Cleanup
handle_worker()in preparation for #2815 (Stop generation early) (#3573)" by @yk in #3644 - To support pip install for model_training package. by @zhanglu0704 in #3643
- Added plugin repositories to add-plugin menu by @yk in #3638
- Add paper acknowledgement form by @yk in #3637
- Add rope-scaling parameters to export_model.py by @andreaskoepf in #3618
- Bump eslint from 8.41.0 to 8.46.0 in /website by @dependabot in #3642
- Bump react-hook-form from 7.44.3 to 7.45.4 in /website by @dependabot in #3641
- Web App Manifest - Save as desktop or mobile app by @sun-jay in #3632
- Long context datasets by @shahules786 in #3646
- Make sure eval subset is sampled without replacing by @andreaskoepf in #3651
- Add loader for megacode2 (rombodawg/LosslessMegaCodeTrainingV2_1m_Evol_Uncensored) by @andreaskoepf in #3650
- Add pretokenizer utility by @andreaskoepf in #3654
- Add megacode3 dataset by @andreaskoepf in #3656
- fix: unsupported message typo by @0o001 in #3658
- Add dataset adapter for loading dolly15k_multilingual dataset by @andreaskoepf in #3660
- More translations into german by @Lo10Th in #3662
- Support ChatML prompt format in worker by @olliestanley in #3668
New Contributors
- @zhanglu0704 made their first contribution in #3643
- @sun-jay made their first contribution in #3632
- @0o001 made their first contribution in #3658
Full Changelog: v0.0.3-alpha36...v0.0.3-alpha37