For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /get-started.md).
Mojo package
nvidia
NVIDIA GPU attention kernels and tile-scheduling utilities.
Architecture-specific implementations live under sm90/ (Hopper) and sm100/ (Blackwell).
Packages
-
sm100: NVIDIA SM100 (Blackwell) attention kernels. -
sm90: NVIDIA SM90 (Hopper) flash-attention v3 kernels and utilities.
Modules
-
common: Shared NVIDIA GPU attention primitives used by both SM90 and SM100 kernels. -
mha_tile_scheduler: Provides tile schedulers for multi-head attention kernels on NVIDIA GPUs.