🐳 Efficient Triton implementations for "Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention"

Implementation of the sparse attention pattern proposed by the Deepseek team in their "Native Sparse Attention" paper

🚀🚀 Efficient implementations of Native Sparse Attention

Efficient triton implementation of Native Sparse Attention.

An efficient implementation of the NSA (Native Sparse Attention) kernel

DeepSeek Native Sparse Attention pytorch implementation

qwen-nsa

VideoNSA: Native Sparse Attention Scales Video Understanding

research impl of Native Sparse Attention (2502.11089)

No description available

No description available

Native Sparse Attention PyTorch Implementaion

native sparse attention kernel

Efficient CUDA implementation of "Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention"

Native Sparse Attention (NSA) M0 steel-thread: SDPA-only, RoPE, CSR mapping, batched prefill, masked varlen SDPA, selection packing, tests and CI.

No description available

Code for the paper "Natively Trainable Sparse Attention for Hierarchical Point Cloud Datasets"

PyTorch implementation of Native Sparse Attention.

For Native Sparse Attention

implementations of native sparse attention (triton) to some models

Implementation of the Proposed Method for Deep Seek Native Sparse Attention (NSA, arXiv 2502.11089)

No description available

Building Native Sparse Attention

Implementation of the Proposed Method for Deep Seek Native Sparse Attention (NSA, arXiv 2502.11089)

No description available

an implementastion of native sparse attention by deepseek on the Mac OS

Implementations for "Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention"

No description available

No description available

No description available