fix(vllm_strategy): llm.generate params 'prompt_token_ids' become part of 'prompts'

[huang3eng]

• bug: TypeError: LLM.generate() got an unexpected keyword argument 'prompt_token_ids', when using LLMJudgeRewardWorker on vllm 0.10.2.

• fix: The _run_local_inference method of LLMJudgeRewardWorker will call llm.generate, and the vllm llm.generate interface has modified the input parameters. The parameter prompt_token_ids has been deprecated, and prompt_token_ids is now part of the prompts parameter.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

benyi seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}