Fix vLLM server-mode communicator initialization to use the current A…

[sywangyi]

…ccelerator device instead of always querying CUDA.

What does this PR do?

Fixes # (issue)

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline, Pull Request section?
• Was this discussed/approved via a GitHub issue? Please add a link to it if that's the case.
• Did you make sure to update the documentation with your changes?
• Did you write any new necessary tests?

AI writing disclosure

We welcome the use of AI tools to help with contributions. For transparency and to help us improve our review process, please indicate the level of AI involvement in this PR.

• No AI usage: the PR was written entirely by a human.
• AI-assisted: some parts were suggested or improved by AI, but the PR was written and reviewed by a human.
• AI-generated: the PR was mostly or fully generated by an AI tool.

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR.

---

Note

Low Risk
One-line device argument change in server-mode initialization only; no auth, data, or broad behavioral refactors.

Overview
In vLLM server mode, weight-sync communicator setup now passes accelerator.device into VLLMClient.init_communicator instead of torch.cuda.current_device().

That matches the device Accelerate assigned for the training main process, so server-mode weight updates work on non-CUDA backends (and stay correct when CUDA index and Accelerate’s device differ).

^{Reviewed by Cursor Bugbot for commit a78ca1a. Bugbot is set up for automated code reviews on this repo. Configure here.}

[sywangyi]

@qgallouedec please help review

[sywangyi]

@IlyasMoutawwakil please help review

[IlyasMoutawwakil]

LGTM

[IlyasMoutawwakil]

@qgallouedec please merge when you can 🙏