GitLab

* [chatgpt]fix train_rm bug with lora

* [chatgpt]support colossalai strategy to train rm

* fix pre-commit

* fix pre-commit 2