[HollowMan6]

What does this PR do?

from vllm-project/vllm@aafd4d2

vLLM changed WorkerWrapperBase signature (i.e., removing vllm_config from __init__). We keep a small runtime fallback to support multiple versions.

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, veomni, sglang, vllm, rollout, trainer, ci, training_utils, recipe, hardware, deployment, ray, worker, single_controller, misc, perf, model, algo, env, tool, ckpt, doc, data, cfg, reward
  • 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.

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).)
• If your PR is related to the recipe submodule, please also update the reference to the submodule commit via git submodule update --remote or cd recipe && git pull origin main.

[gemini-code-assist]

Code Review

This pull request addresses a breaking change in a newer version of vllm where the vllm_config argument was removed from the WorkerWrapperBase initializer. The fix introduces a new method, _build_inference_engine, which uses a try-except block to handle both old and new versions of vllm. This is a good approach for ensuring compatibility. My review includes a suggestion to make this compatibility check more robust by using Python's inspect module instead of catching a broad TypeError, which will prevent masking other potential errors.

[Copilot]

Pull request overview

This PR fixes a compatibility issue with newer vLLM versions where the WorkerWrapperBase.__init__ signature changed by removing the vllm_config parameter. The solution implements a runtime fallback pattern using try-except to support both old and new vLLM versions.

Changes:

• Added a new _build_inference_engine() helper method that attempts to create WorkerWrapperBase with vllm_config and falls back to no-argument initialization for newer versions
• Replaced direct instantiation of WorkerWrapperBase with a call to the new helper method

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[gemini-code-assist]

Code Review

This pull request addresses a compatibility issue with newer versions of vLLM by conditionally passing the vllm_config argument to WorkerWrapperBase based on its __init__ signature. The approach is sound and correctly fixes the problem. My review includes a suggestion to refactor the implementation to use a try-except block, which is a more robust and idiomatic Python pattern for this kind of version compatibility check. This change would also remove the need for an extra import.

[Copilot]

Pull request overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated no new comments.

Comments suppressed due to low confidence (1)

verl/workers/rollout/vllm_rollout/vllm_rollout.py:212

• The self.vllm_config is assigned after the method call to _build_inference_engine() at line 228, but _build_inference_engine() attempts to access self.vllm_config at line 190. This ordering issue means that when newer vLLM versions don't require vllm_config the code works fine, but for older versions this could cause an AttributeError if self.vllm_config hasn't been set yet. The _build_inference_engine() call should be moved after line 212 where self.vllm_config is assigned.

        self.vllm_config = all_kwargs[0]["vllm_config"]

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.