Skip to content

Keep returns separate from advantages in GRPO #773

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

Open
ebronstein wants to merge 2 commits into
base: main
Choose a base branch
from
+3 −1
Open

Keep returns separate from advantages in GRPO #773

Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6f5afbe
Keep returns separate from advantages in GRPO.
ebronstein Dec 11, 2025
c6f3c4e
Make returns the correct shape and mask with response mask.
ebronstein Dec 11, 2025
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
3 changes: 2 additions & 1 deletion skyrl-train/skyrl_train/utils/ppo_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1051,6 +1051,7 @@ def compute_grpo_outcome_advantage(
"""
# this assumes response-level rewards
scores = token_level_rewards.sum(dim=-1)
returns = scores.clone()

id2score = defaultdict(list)
id2mean = {}
Expand All @@ -1076,7 +1077,7 @@ def compute_grpo_outcome_advantage(
scores[i] = scores[i] - id2mean[index[i]]
scores = scores.unsqueeze(-1) * response_mask

return scores, scores
return scores, returns


def repopulate_all_registries():
Expand Down
Loading