Added error check to RLOO, PPOv2, OnlineDPO that `ref_policy` and `policy` have different identities#2057

Merged

qgallouedec merged 9 commits intohuggingface:mainfrom RylanSchaeffer:main

Sep 17, 2024

+119-1

Commits on Sep 11, 2024

Added error check to RLOO, PPOv2, OnlineDPO that ref_policy and policy should have different identities.
RylanSchaeffer
committed

Commits on Sep 13, 2024

Update online_dpo_trainer.py

qgallouedec
and
lewtun
authored