🐳 Efficient Triton implementations for "Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention"
☆1,001Feb 5, 2026Updated 3 months ago
Alternatives and similar repositories for native-sparse-attention
Users that are interested in native-sparse-attention are comparing it to the libraries listed below. We may earn a commission when you buy through links labeled 'Ad' on this page.
Sorting:
- Implementation of the sparse attention pattern proposed by the Deepseek team in their "Native Sparse Attention" paper☆805Aug 15, 2025Updated 9 months ago
- Efficient triton implementation of Native Sparse Attention.☆277May 23, 2025Updated last year
- 🚀 Efficient implementations for emerging model architectures☆5,116Updated this week
- MoBA: Mixture of Block Attention for Long-Context LLMs☆2,120Apr 3, 2025Updated last year
- An efficient implementation of the NSA (Native Sparse Attention) kernel☆133Jun 24, 2025Updated 10 months ago
- Deploy on Railway without the complexity - Free Credits Offer • AdConnect your repo and Railway handles the rest with instant previews. Quickly provision container image services, databases, and storage volumes.
- A sparse attention kernel supporting mix sparse patterns☆516Jan 18, 2026Updated 4 months ago
- qwen-nsa☆87Oct 14, 2025Updated 7 months ago
- Muon is Scalable for LLM Training☆1,480Aug 3, 2025Updated 9 months ago
- Distributed Compiler based on Triton for Parallel Systems☆1,426Apr 22, 2026Updated last month
- Quantized Attention on GPU☆44Nov 22, 2024Updated last year
- Helpful tools and examples for working with flex-attention☆1,187Apr 13, 2026Updated last month
- FlashInfer: Kernel Library for LLM Serving☆5,621May 16, 2026Updated last week
- [ICLR 2025] DuoAttention: Efficient Long-Context LLM Inference with Retrieval and Streaming Heads☆541Feb 10, 2025Updated last year
- A Distributed Attention Towards Linear Scalability for Ultra-Long Context, Heterogeneous Data Training☆819Updated this week
- Virtual machines for every use case on DigitalOcean • AdGet dependable uptime with 99.99% SLA, simple security tools, and predictable monthly pricing with DigitalOcean's virtual machines, called Droplets.
- [ICML2025] SpargeAttention: A training-free sparse attention that accelerates any model inference.☆995Feb 25, 2026Updated 2 months ago
- 🔥 A minimal training framework for scaling FLA models☆389Apr 22, 2026Updated last month
- Ring attention implementation with flash attention☆1,020Sep 10, 2025Updated 8 months ago
- [NeurIPS'24 Spotlight, ICLR'25, ICML'25] To speed up Long-context LLMs' inference, approximate and dynamic sparse calculate the attention…☆1,212Apr 8, 2026Updated last month
- [ICML 2024] Quest: Query-Aware Sparsity for Efficient Long-Context LLM Inference☆384Jul 10, 2025Updated 10 months ago
- Domain-specific language designed to streamline the development of high-performance GPU/CPU/Accelerators kernels☆6,278Updated this week
- ☆116Feb 25, 2025Updated last year
- [MLSys'25] QServe: W4A8KV4 Quantization and System Co-design for Efficient LLM Serving; [MLSys'25] LServe: Efficient Long-sequence LLM Se…☆838Mar 6, 2025Updated last year
- DeeperGEMM: crazy optimized version☆86May 5, 2025Updated last year
- Serverless GPU API endpoints on Runpod - Get Bonus Credits • AdSkip the infrastructure headaches. Auto-scaling, pay-as-you-go, no-ops approach lets you focus on innovating your application.
- DeepSeek Native Sparse Attention pytorch implementation☆117Dec 17, 2025Updated 5 months ago
- ☆66Apr 26, 2025Updated last year
- [ICLR 2025] COAT: Compressing Optimizer States and Activation for Memory-Efficient FP8 Training☆263Aug 9, 2025Updated 9 months ago
- Efficient Triton Kernels for LLM Training☆6,365Updated this week
- Code for paper: [ICLR2025 Oral] FlexPrefill: A Context-Aware Sparse Attention Mechanism for Efficient Long-Sequence Inference☆168Oct 13, 2025Updated 7 months ago
- ☆248Nov 19, 2025Updated 6 months ago
- ☆52May 19, 2025Updated last year
- A fast communication-overlapping library for tensor/expert parallelism on GPUs.☆1,309Aug 28, 2025Updated 8 months ago
- USP: Unified (a.k.a. Hybrid, 2D) Sequence Parallel Attention for Long Context Transformers Model Training and Inference☆667Jan 15, 2026Updated 4 months ago
- Virtual machines for every use case on DigitalOcean • AdGet dependable uptime with 99.99% SLA, simple security tools, and predictable monthly pricing with DigitalOcean's virtual machines, called Droplets.
- Tile primitives for speedy kernels☆3,360May 11, 2026Updated last week
- BitBLAS is a library to support mixed-precision matrix multiplications, especially for quantized LLM deployment.☆764Aug 6, 2025Updated 9 months ago
- flash attention tutorial written in python, triton, cuda, cutlass☆516Jan 20, 2026Updated 4 months ago
- Efficient LLM Inference over Long Sequences☆395Jun 25, 2025Updated 10 months ago
- Fast low-bit matmul kernels in Triton☆454May 15, 2026Updated last week
- DeepGEMM: clean and efficient FP8 GEMM kernels with fine-grained scaling☆7,272May 13, 2026Updated last week
- A bidirectional pipeline parallelism algorithm for computation-communication overlap in DeepSeek V3/R1 training.☆2,953Jan 14, 2026Updated 4 months ago