[Perf] Optimize Preparing Inputs for GPU Model Runner

[SnowCharmQ]

This PR enhances the performance of the method _prepare_inputs in gpu_model_runner.py by replacing the original Python loop implementation with map and numpy array operations.

On my clusters, it can achieve nearly a twofold performance improvement.

[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 @SnowCharmQ, this is great!

On my clusters, it can achieve nearly a twofold performance improvement.

Presumably you're referring to the improvement of this loop, not end-to-end? :)

[SnowCharmQ]

Thanks @SnowCharmQ, this is great!

On my clusters, it can achieve nearly a twofold performance improvement.

Presumably you're referring to the improvement of this loop, not end-to-end? :)

Hi @njhill, the improvement refers to the loop exactly. Sorry for the confusion :)

[SnowCharmQ]

Hi @njhill, I noticed an issue with the CI check. Do you have any idea what might be going wrong and how it can be resolved?

[DarkLight1337]

I retried the test and it passes now