IMPORTANT: To view this page as Markdown, append `.md` to the URL (e.g. /max/get-started.md). For the complete documentation index, see llms.txt.
Skip to main content
For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).

Mojo module

mha

SM90 (Hopper) FlashAttention-3 multi-head attention kernel and dispatch layer.

This module implements the warp-specialized, TMA-based MHA kernel for NVIDIA H100 GPUs, along with the dispatch chain that materializes comptime configuration choices (scheduler, sink, KV row offsets, ragged valid lengths) into concrete kernel instantiations.

Functions

Was this page helpful?