Fix DPO and RLOO incompatibility with FSDP2#4838

qgallouedec merged 7 commits intohuggingface:mainhuggingface/trl:mainfrom

flutist:make_dpo_and_fsdp2_successflutist/trl_dpo_compatible_qwen3vl:make_dpo_and_fsdp2_successCopy head branch name to clipboard