v0.0.6

What's Changed

• Add length-regularized DPO by @lkevinzc in #24
• Refactor and add PPO for math reasoning by @lkevinzc in #25

Full Changelog: v0.0.5...v0.0.6

v0.0.5

What's Changed

• fix typo by @lkevinzc in #13
• Support async online training and offline training by @lkevinzc in #15
• support LR_DPO by @cameron-chen in #17
• Add BNF Loss by @MaoXinn in #18
• Add more benchmarking scripts by @lkevinzc in #19
• Update BNF by @MaoXinn in #20
• Fix typos and formatting by @lkevinzc in #21
• revert lr dpo by @lkevinzc in #22
• bump version by @lkevinzc in #23

New Contributors

• @cameron-chen made their first contribution in #17
• @MaoXinn made their first contribution in #18

Full Changelog: v0.0.4...v0.0.5

v0.0.4

What's Changed

• Fix image width by @lkevinzc in #8
• Update README.md by @lkevinzc in #9
• clearer readme examples & rename reward_oracle to preference_oracle by @lkevinzc in #11
• dump version 0.0.4 by @lkevinzc in #12

Full Changelog: v0.0.3...v0.0.4

v0.0.3

What's Changed

• Refactor arg parsing and add examples by @lkevinzc in #6
• Fix APL due to vllm upgrade; update package v0.0.3 by @lkevinzc in #7
• Release paper on Arxiv: https://arxiv.org/pdf/2411.01493

Full Changelog: v0.0.2...v0.0.3

The first OSS release

v0.0.2

dump version (#5)