[rollout,vllm] feat: disable sleep mode in fully-async mode

[chenjiaoAngel]

What does this PR do?

Add concise overview of what this PR aims to achieve or accomplish. Reference related GitHub issues and PRs that help with the review.

• 在fully-async 模式下，添加enable_sleep_mode 参数支持，可以使用虚拟显存功能，可充分利用显存，减少OOM 现象
• 默认下是True，与原来保持一致

Checklist Before Starting

• Search for similar PRs. Paste at least one query link here: ...
• Format the PR title as [{modules}] {type}: {description} (This will be checked by the CI)
  • {modules} include fsdp, megatron, sglang, vllm, rollout, trainer, ci, training_utils, recipe, hardware, deployment, ray, worker, single_controller, misc, perf, model, algo, env, tool, ckpt, doc, data
  • If this PR involves multiple modules, separate them with , like [megatron, fsdp, doc]
  • {type} is in feat, fix, refactor, chore, test
  • If this PR breaks any API (CLI arguments, config, function signature, etc.), add [BREAKING] to the beginning of the title.
  • Example: [BREAKING][fsdp, megatron] feat: dynamic batching

Test

For changes that can not be tested by CI (e.g., algorithm implementation, new model support), validate by experiment(s) and show results like training curve plots, evaluation results, etc.

可参考[recipe/fully_async_policy/shell/dapo_30b_a3b_base_math_fsdp.sh] 脚本进行使用，新增+actor_rollout_ref.rollout.enable_sleep_mode=False\ 命令即可；

API and Usage Example

Demonstrate how the API changes if any, and provide usage example(s) if possible.

# Add code snippet or script demonstrating how to use this

Design & Code Changes

Demonstrate the high-level design if this PR is complex, and list the specific changes.

Checklist Before Submitting

Important

Please check all the following items before requesting a review, otherwise the reviewer might deprioritize this PR for review.

• Read the Contribute Guide.
• Apply pre-commit checks: pre-commit install && pre-commit run --all-files --show-diff-on-failure --color=always
• Add / Update the documentation.
• Add unit or end-to-end test(s) to the CI workflow to cover all the code. If not feasible, explain why: ...
• Once your PR is ready for CI, send a message in the ci-request channel in the verl Slack workspace. (If not accessible, please try the Feishu group (飞书群).)

[gemini-code-assist]

Code Review

This pull request introduces an enable_sleep_mode flag to control a vLLM server setting, which in turn affects CUDA memory allocation behavior. The changes span a new shell script for configuration, a modification to the rollout configuration class, and logic in the vLLM async server. My review identifies a critical bug in the new shell script that would cause it to fail, and a high-severity logging inconsistency in the Python code. I've provided suggestions to fix both issues.

[wuxibin89]

Please format code: https://github.com/volcengine/verl/blob/main/CONTRIBUTING.md#code-linting-and-formatting

[ArronHZG]

Does this PR suggest that turning off sleep mode during rollout can save GPU memory? If so, approximately how much can it save?

[chenjiaoAngel]

Does this PR suggest that turning off sleep mode during rollout can save GPU memory? If so, approximately how much can it save?

in my case, it can save at least 2-3G; and it also can improve profile

[chenjiaoAngel]

Please format code: https://github.com/volcengine/verl/blob/main/CONTRIBUTING.md#code-linting-and-formatting

已更新～