Online DPO Support
Open
community-request
enhancement
- Dominant language
- Python
- Stars
- 2k
- Forks
- 561
- Avg merge
- 4d 5h
- Merged PRs (30d)
- 145
Description
Hey Folks,
Does Nemo-RL support Online DPO(https://arxiv.org/abs/2402.04792)?
Given there are[ environments](https://docs.nvidia.com/nemo/rl/latest/guides/environments.html) in Nemo-RL, was wondering about the possibility of generating preferences online based on the outcomes in the environment and fine-tuning.
Contributor guide
Assessment
This issue has not been assessed yet.