Optimize input preparation for FlashInfer [2/N]

[WoosukKwon]

Should be merged after #23147

[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.

🚀

[gemini-code-assist]

Code Review

This pull request introduces significant optimizations to the input preparation logic for the FlashInfer attention backend. The key changes include refactoring metadata handling by moving static parameters from FlashInferMetadata to FlashInferMetadataBuilder, replacing CPU-intensive PyTorch operations with faster NumPy equivalents, and substituting a slow torch.masked_select with a custom Triton kernel for preparing paged KV indices. Additionally, the calculation of max_seq_len is optimized by pre-computing it once. These changes are well-implemented and should lead to noticeable performance improvements. The code quality is high, and the optimizations are sound.

[mergify]

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @WoosukKwon.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

[mergify]

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @WoosukKwon.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

[nvpohanh]

@WoosukKwon We found that this optimization can reduce gaps between decoding steps when running with low concurrency. Do you plan to continue working on this so that this can be merged? Thanks!