tilde-research / nsa-implLinks
An efficient implementation of the NSA (Native Sparse Attention) kernel
☆110Updated last month
Alternatives and similar repositories for nsa-impl
Users that are interested in nsa-impl are comparing it to the libraries listed below
Sorting:
- ☆123Updated 2 months ago
- ☆232Updated 2 months ago
- ☆79Updated 5 months ago
- Efficient triton implementation of Native Sparse Attention.☆186Updated 2 months ago
- [ICML 2025] SparseLoRA: Accelerating LLM Fine-Tuning with Contextual Sparsity☆49Updated last month
- Flash-Muon: An Efficient Implementation of Muon Optimizer☆152Updated last month
- [ICML 2025] XAttention: Block Sparse Attention with Antidiagonal Scoring☆213Updated last month
- The evaluation framework for training-free sparse attention in LLMs☆88Updated last month
- Accelerate LLM preference tuning via prefix sharing with a single line of code☆42Updated last month
- ☆137Updated 5 months ago
- ☆54Updated last month
- Code for paper: [ICLR2025 Oral] FlexPrefill: A Context-Aware Sparse Attention Mechanism for Efficient Long-Sequence Inference☆124Updated 2 months ago
- 16-fold memory access reduction with nearly no loss☆103Updated 4 months ago
- [CoLM'25] The official implementation of the paper <MoA: Mixture of Sparse Attention for Automatic Large Language Model Compression>☆145Updated 3 weeks ago
- A curated list of recent papers on efficient video attention for video diffusion models, including sparsification, quantization, and cach…☆32Updated 2 weeks ago
- Fast and memory-efficient exact attention☆69Updated 5 months ago
- Beyond KV Caching: Shared Attention for Efficient LLMs☆19Updated last year
- 🔥 A minimal training framework for scaling FLA models☆220Updated last month
- Quantized Attention on GPU☆44Updated 8 months ago
- ☆51Updated last month
- Odysseus: Playground of LLM Sequence Parallelism☆72Updated last year
- ☆81Updated last week
- ☆113Updated 2 months ago
- HALO: Hadamard-Assisted Low-Precision Optimization and Training method for finetuning LLMs. 🚀 The official implementation of https://arx…☆18Updated 5 months ago
- GEAR: An Efficient KV Cache Compression Recipefor Near-Lossless Generative Inference of LLM☆165Updated last year
- ☆75Updated 2 months ago
- 🎬 3.7× faster video generation E2E 🖼️ 1.6× faster image generation E2E ⚡ ColumnSparseAttn 9.3× vs FlashAttn‑3 💨 ColumnSparseGEMM 2.5× …☆78Updated last month
- Kinetics: Rethinking Test-Time Scaling Laws☆70Updated 3 weeks ago
- ☆50Updated 2 months ago
- Transformers components but in Triton☆34Updated 2 months ago