vipshop/ffpa-attn

Fast and memory-efficient exact attention

FlashMLA: Efficient Multi-head Latent Attention Kernels

CUDA kernels for linear attention variants, written in CuTe DSL and CUTLASS C++.

🚀🚀 Efficient implementations of Native Sparse Attention

Helpful tools and examples for working with flex-attention

high-performance linear attention kernel library built on TileLang

FlashKDA: high-performance Kimi Delta Attention kernels

Trainable fast and memory-efficient sparse attention

cuDNN Frontend is NVIDIA's modern, open-source entry point to the cuDNN library and a growing collection of high-performance open-source kernels.