Skip to content

[recipe] fix: Fixed scripts for one_step_off_policy async not implemention #4350

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

Merged
FightingZhen merged 1 commit into from
Nov 29, 2025
Merged

[recipe] fix: Fixed scripts for one_step_off_policy async not implemention #4350

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
1 change: 1 addition & 0 deletions recipe/one_step_off_policy/grpo_qwen3_8b_gsm8k_fsdp2_8_8_npu.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -60,6 +60,7 @@ python3 -m recipe.one_step_off_policy.main_ppo \
actor_rollout_ref.rollout.log_prob_micro_batch_size_per_gpu=1 \
actor_rollout_ref.rollout.tensor_model_parallel_size=4 \
actor_rollout_ref.rollout.name=vllm \
actor_rollout_ref.rollout.mode=sync \
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist bot Nov 29, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

high

This change correctly sets the rollout mode to sync. The script uses the recipe.one_step_off_policy.main_ppo entry point, which seems intended for synchronous execution, similar to what's done in the E2E test script tests/special_e2e/run_one_step_off_policy.sh. Without this explicit setting, it would default to async mode from the configuration file verl/trainer/config/rollout/rollout.yaml, which is likely incorrect for this script and could lead to unexpected behavior. This ensures the script runs in its intended synchronous mode.

actor_rollout_ref.rollout.gpu_memory_utilization=0.8 \
actor_rollout_ref.rollout.max_num_batched_tokens=$((max_prompt_length + max_response_length)) \
actor_rollout_ref.rollout.n=8 \
Expand Down
Loading