NVIDIA / NVIDIA/TensorRT-LLM

[Feature]: Enable MLA prefill-only operator

Open
#8,424 0 comments 0 reactions 1 assignee View on GitHub

@galagam is already working on this.

Since Jan 7, 2026.

AutoDeploy Customized kernels
Dominant language
Python
Stars
14.7k
Forks
2.8k
Avg merge
2d 23h
Merged PRs (30d)
489

Description

🚀 The feature, motivation and pitch

FlashInfer MLA MHA-mode (no weight absorption) with cache; ragged; for prefill-only. Uses flashinfer.BatchPrefillWithRaggedKVCacheWrapper

This kernel is optimized for prefill.
Its inputs are full heads Q/K/V so it performs MLA in MHA-mode (no absorption).
The operator should compute ckv and kpe and write them to the compressed cache, then continue to do MHA.
This can probably done with other flashinfer kernels.

Alternatives

No response

Additional context

No response

Before submitting a new issue...
  • Make sure you already searched for relevant issues, and checked the documentation and examples for answers to frequently asked questions.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.