Skip to content

[rollout, vllm, sglang] fix: allow user customization of repetition_penalty to avoid watchdog timeout during GRPO rollout #3309

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
merged 6 commits into from
Sep 5, 2025
Merged

[rollout, vllm, sglang] fix: allow user customization of repetition_penalty to avoid watchdog timeout during GRPO rollout #3309

Changes from 1 commit
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
bcaeaf6
[rollout, vllm] fix: allow user customization of `repetition_penalty`…
Mighten Sep 2, 2025
bb9e2c7
put user-specified `repetition_penalty` to vllm async server
Mighten Sep 2, 2025
6566dba
put user-specified `repetition_penalty` to sglang rollout engine
Mighten Sep 2, 2025
481eacb
put user-specified `repetition_penalty` to sglang async server
Mighten Sep 2, 2025
1b83e3c
prioritize user-provided `repetition_penalty` over the default value …
Mighten Sep 3, 2025
6418a63
Merge branch 'refs/heads/main' into vllm-rollout-repetition_penalty
Mighten Sep 4, 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
1 change: 1 addition & 0 deletions verl/workers/rollout/vllm_rollout/vllm_rollout_spmd.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -208,6 +208,7 @@ def __init__(self, model_path: str, config: RolloutConfig, tokenizer, model_hf_c
n=1,
logprobs=0, # can be set to 0 and let actor to recompute
max_tokens=config.response_length,
repetition_penalty=config.get("repetition_penalty", 1.0),
)

kwargs["detokenize"] = False
Expand Down