v0.0.2
版本发布时间: 2023-12-17 23:17:11
OpenRLHF/OpenRLHF最新发布版本:v0.4.2(2024-08-29 19:39:57)
Changes
- Remove pad_token for llama2
- Support cDPO/IPO
- Fix Ray RLHF sync bugs
- Optimized eos_indicies with
torch.argmax - Fix local datasets
- Fix DPO DataLoader bugs
MyGit