An efficient implementation of the NSA (Native Sparse Attention) kernel
☆129Jun 24, 2025Updated 8 months ago
Alternatives and similar repositories for nsa-impl
Users that are interested in nsa-impl are comparing it to the libraries listed below
Sorting:
- ☆15Mar 2, 2025Updated last year
- ☆19Dec 4, 2025Updated 2 months ago
- FlexAttention w/ FlashAttention3 Support☆27Oct 5, 2024Updated last year
- 🐳 Efficient Triton implementations for "Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention"☆969Feb 5, 2026Updated 3 weeks ago
- Benchmark tests supporting the TiledCUDA library.☆18Nov 19, 2024Updated last year
- Less Is More: Training-Free Sparse Attention with Global Locality for Efficient Reasoning☆29Sep 12, 2025Updated 5 months ago
- Efficient triton implementation of Native Sparse Attention.☆268May 23, 2025Updated 9 months ago
- DeeperGEMM: crazy optimized version☆74May 5, 2025Updated 9 months ago
- The simplest implementation of recent Sparse Attention patterns for efficient LLM inference.☆91Jul 17, 2025Updated 7 months ago
- ☆20May 30, 2024Updated last year
- ☆134May 29, 2025Updated 9 months ago
- Code for NeurIPS 2024 Spotlight: "Scaling Laws and Compute-Optimal Training Beyond Fixed Training Durations"☆92Oct 30, 2024Updated last year
- [ICLR 2025 & COLM 2025] Official PyTorch implementation of the Forgetting Transformer and Adaptive Computation Pruning☆140Updated this week
- a simple API to use CUPTI☆11Aug 19, 2025Updated 6 months ago
- A method for evaluating the high-level coherence of machine-generated texts. Identifies high-level coherence issues in transformer-based …☆11Mar 18, 2023Updated 2 years ago
- Unofficial implementation of paper : Exploring the Space of Key-Value-Query Models with Intention☆12May 24, 2023Updated 2 years ago
- ☆11Oct 11, 2023Updated 2 years ago
- mHC-lite: You Don’t Need 20 Sinkhorn-Knopp Iterations☆70Jan 12, 2026Updated last month
- Awesome Triton Resources☆39Apr 27, 2025Updated 10 months ago
- Code for the paper "Function-Space Learning Rates"☆25Jun 3, 2025Updated 9 months ago
- Medusa: Accelerating Serverless LLM Inference with Materialization [ASPLOS'25]☆41May 13, 2025Updated 9 months ago
- ☆272Jun 6, 2025Updated 8 months ago
- Transformers components but in Triton☆34May 9, 2025Updated 9 months ago
- Tritonbench is a collection of PyTorch custom operators with example inputs to measure their performance.☆327Updated this week
- Code for the paper "Stack Attention: Improving the Ability of Transformers to Model Hierarchical Patterns"☆18Mar 15, 2024Updated last year
- Cute layout visualization☆30Jan 18, 2026Updated last month
- ☆15Feb 24, 2026Updated last week
- Combining SOAP and MUON☆19Feb 11, 2025Updated last year
- Fast and memory-efficient exact attention☆18Feb 23, 2026Updated last week
- Flash-Muon: An Efficient Implementation of Muon Optimizer☆237Jun 15, 2025Updated 8 months ago
- ☆23Jun 18, 2024Updated last year
- Framework to reduce autotune overhead to zero for well known deployments.☆97Sep 19, 2025Updated 5 months ago
- Triton-based implementation of Sparse Mixture of Experts.☆266Oct 3, 2025Updated 5 months ago
- Tilus is a tile-level kernel programming language with explicit control over shared memory and registers.☆446Feb 4, 2026Updated 3 weeks ago
- NVIDIA cuTile learn☆163Dec 9, 2025Updated 2 months ago
- [NeurIPS 2025, Spotlight]: Ambient-o: Training Good models with Bad Data.☆30Jan 21, 2026Updated last month
- Implementation from scratch in C of the Multi-head latent attention used in the Deepseek-v3 technical paper.☆18Jan 15, 2025Updated last year
- [EMNLP 2023] Official implementation of the algorithm ETSC: Exact Toeplitz-to-SSM Conversion our EMNLP 2023 paper - Accelerating Toeplitz…☆14Oct 17, 2023Updated 2 years ago
- Reference implementation of "Softmax Attention with Constant Cost per Token" (Heinsen, 2024)☆24Jun 6, 2024Updated last year