[vllm, rollout] feat: support reset prefix cache after abort

[PeterSH6]

What does this PR do?

Add a reset option after calling the abort request in vLLM.
Default reset prefix cache after abort.
If not reset, it may become pipelinerl.

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.

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 a reset_prefix_cache parameter to allow resetting the prefix cache after aborting requests. While this is a good addition, there is a critical issue that needs to be addressed: the new parameter is not propagated through the vLLMReplica methods (abort_all_requests and abort_request in verl/workers/rollout/vllm_rollout/vllm_async_server.py). This makes the feature non-functional for callers of vLLMReplica. Please update these methods to accept and pass the reset_prefix_cache parameter to the underlying server calls. Additionally, I've noted an opportunity to improve maintainability by refactoring duplicated code for resetting the cache.

[gemini-code-assist]

This block of code for resetting the prefix cache is also present in the abort_request method (lines 598-604). This code duplication goes against the DRY (Don't Repeat Yourself) principle and can lead to maintenance issues where one instance is updated but the other is missed. I recommend extracting this logic into a private helper method, for instance _try_reset_prefix_cache, and calling it from both abort_all_requests and abort_request.

[wuxibin89]

We should always reset_prefix_cache when abort request?

[PeterSH6]

Yes, we should reset prefix cache after abort as the rollout weights are usually updated after aborting.

[PeterSH6]

If not reset, it may degenerate to a paradigm similar to pipelineRL

[wuxibin89]

When will we want to abort single request?

[PeterSH6]

This may not be used quite often, but we can leave it as an API for fine-grained control