Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

fix DS for peft ref_model in ppo trainer #309

Merged
merged 1 commit into from
Apr 25, 2023
Merged

fix DS for peft ref_model in ppo trainer #309

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion trl/trainer/ppo_trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -281,7 +281,7 @@ def __init__(
)
if is_deepspeed_used:
# 8 bit models are already set on the correct device
if not getattr(self.ref_model.pretrained_model, "is_loaded_in_8bit", False):
if not self.is_peft_model and not getattr(self.ref_model.pretrained_model, "is_loaded_in_8bit", False):
# DS integration only allows for single model and as `ref_model` is only used for
# `KL devergence loss`,i.e, in eval model, just have it be on the respective device and
# there is no need to pass it to the `accelerator.prepare` call
Expand Down