What's Changed

No pull requests were merged in this release.

Full Changelog: v0.1.13...v0.1.14

What's Changed

• refactor: Make W&B API key optional (#43)
• fix delete ckpt bug that errored out the history_df step (#42)
• Update notebooks (#41)
• Update notebooks (#40)
• Add alpha message (#39)
• Update discord link (#38)
• Add link to blog post (#37)
• Clean up 2048 notebook (#36)

Full Changelog: v0.1.12...v0.1.13

What's Changed

No pull requests were merged in this release.

Full Changelog: v0.1.10...v0.1.12

What's Changed

No pull requests were merged in this release.

Full Changelog: v0.1.9...v0.1.10

What's Changed

• refactor: Unsloth backend improvements (#5)
• refactor: Unsloth architecture/configs (#8)
• refactor: Model project (#31)
• refactor: Introduce ClientWrapper for backwards compatibility (#7)
• refactor: Deprecate gather_groups, introduce gather_trajectories (#3)
• refactor: API changes (#27)
• perf: Performance improvements (#15)
• feature: Add Unsloth KL divergence support (#11)
• feat: Add async generator support to mp_actors package (#12)
• feat: Add Unsloth progress bar (#13)
• chore: Remove tools (#22)
• chore: Large batch HN title generator variant (#16)
• chore: Clean up unused code (#21)
• [WIP] Local api fixes (#6)
• [WIP] Add tic tac toe example (#4)
• Use a public reward model URL for the hn title generator examples (#18)
• Update license (#20)
• Update 2048 (#24)
• Ttt unsloth (#19)
• Record reward (#14)
• Notebooks (#33)
• Improve README start, clean up notebooks (#28)
• Hn title generator (#17)
• Fix notebook links (#35)
• Allow deps to be installed on mac (#1)
• Add reward standard deviation reporting (#10)
• Add explanations to the 2048 notebook (#30)
• Add docs skeleton (#29)
• Add credits (#34)
• Add code to log completions as OpenPipe request logs (#2)
• Add CONTRIBUTING.md (#32)

Full Changelog: ...v0.1.9