[bugfix] torch profiler bug for single gpu with GPUExecutor

[SolitaryThinker]

GPUExecutor has a different API and does not define a _run_workers.

Another way to fix this would be define the _run_workers (it would only call the driver_worker) api in GPUExecutor to match the API of the other executors. That will avoid the extra import..

closes #8326 and closes #8351

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.
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 do one of these:

• Add ready label to the PR
• Enable auto-merge.

🚀

[SolitaryThinker]

/ready

[njhill]

Different executor class for non-async case

[SolitaryThinker]

Nice catch @njhill! I'm going to use type() to match on GPUExecutor instead, as isinstance will also catch TP > 2 (multiproc) and only call start_profile on the driver_worker instead of all workers.

[SolitaryThinker]

/ready

[SolitaryThinker]

would be good to get a force merge @simon-mo @youkaichao

[DarkLight1337]

Can you investigate whether the failure in LLaVA test is related to your changes?

[2024-09-12T20:14:39Z] INFO 09-12 13:14:39 model_runner.py:997] Starting to load model llava-hf/llava-v1.6-mistral-7b-hf...
[2024-09-12T20:14:39Z] INFO 09-12 13:14:39 weight_utils.py:242] Using model weights format ['*.safetensors']
Loading safetensors checkpoint shards: 100% 4/4 [00:00<00:00, 37.70it/s]
[2024-09-12T20:14:42Z] INFO 09-12 13:14:42 model_runner.py:1008] Loading model weights took 14.0711 GB
[2024-09-12T20:14:42Z] WARNING 09-12 13:14:42 model_runner.py:1176] Computed max_num_seqs (min(256, 10240 // 11712)) to be less than 1. Setting it to the minimum value of 1.
[2024-09-12T20:16:16Z] FAILED