You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
@c-box In arian-askari#1, I have made several modifications to facilitate the upgrade of the transformers version for RL4LMs. While I cannot guarantee that these changes won't introduce any unforeseen issues, I have successfully trained the BLOOM-560M model with batch size 2 on PPO policy for the summarization after this modifications.
The repository uses transformers version 4.18, which does not support bloom, is there any way to use bloom as the initial policy for training?
The text was updated successfully, but these errors were encountered: