[Bugfix] Enable V1 usage stats

[mgoin]

Calling report_usage wasn't ever piped down to V1 LLMEngine or AsyncLLM. @njhill changed accessing the CUDA device information to happen in a subprocess rather than a thread because we run into initializing cuda in the same proc potentially too early.

This could in theory fail if there isn't temporarily enough free gpu mem for the separate cuda context. It wasn't a problem in v0 because it would be called in the driver worker.

Tested manually:

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

🚀

[njhill]

Thanks @mgoin! Could we put this as a util method in v1/utils.py? Looks like it could take just vllm_config and usage_context?

[WoosukKwon]

@simon-mo

[WoosukKwon]

@mgoin Could you please take a look at the failed tests?

[mgoin]

I could not find any clear error with the failing tests other than timeouts and OOMs, so I will rerun to see if it is consistent

[simon-mo]

not sure whether this is related in V1 test failure

[2025-04-23T02:44:35Z] ERROR 04-22 19:44:35 [core.py:461] Invocation of echo method failed
--
  | [2025-04-23T02:44:35Z] ERROR 04-22 19:44:35 [core.py:461] Traceback (most recent call last):
  | [2025-04-23T02:44:35Z] ERROR 04-22 19:44:35 [core.py:461]   File "/usr/local/lib/python3.12/dist-packages/vllm/v1/engine/core.py", line 457, in _handle_client_request
  | [2025-04-23T02:44:35Z] ERROR 04-22 19:44:35 [core.py:461]     method = getattr(self, method_name)
  | [2025-04-23T02:44:35Z] ERROR 04-22 19:44:35 [core.py:461]              ^^^^^^^^^^^^^^^^^^^^^^^^^^
  | [2025-04-23T02:44:35Z] ERROR 04-22 19:44:35 [core.py:461] AttributeError: 'EngineCoreProc' object has no attribute 'echo'
  | [2025-04-23T02:44:35Z] DEBUG 04-22 19:44:35 [core.py:416] EngineCore waiting for work.