PiotrNawrot / nano-sparse-attentionLinks
The simplest implementation of recent Sparse Attention patterns for efficient LLM inference.
☆70Updated last week
Alternatives and similar repositories for nano-sparse-attention
Users that are interested in nano-sparse-attention are comparing it to the libraries listed below
Sorting:
- The source code of our work "Prepacking: A Simple Method for Fast Prefilling and Increased Throughput in Large Language Models" [AISTATS …☆59Updated 8 months ago
- Fast and memory-efficient exact attention☆68Updated 3 months ago
- The evaluation framework for training-free sparse attention in LLMs☆33Updated last week
- Simple and efficient pytorch-native transformer training and inference (batched)☆76Updated last year
- Experiment of using Tangent to autodiff triton☆79Updated last year
- Using FlexAttention to compute attention with different masking patterns☆43Updated 9 months ago
- Language models scale reliably with over-training and on downstream tasks☆97Updated last year
- ☆108Updated last year
- Cold Compress is a hackable, lightweight, and open-source toolkit for creating and benchmarking cache compression methods built on top of…☆137Updated 10 months ago
- ☆81Updated last year
- Tree Attention: Topology-aware Decoding for Long-Context Attention on GPU clusters☆126Updated 6 months ago
- A MAD laboratory to improve AI architecture designs 🧪☆120Updated 6 months ago
- CUDA and Triton implementations of Flash Attention with SoftmaxN.☆70Updated last year
- ☆78Updated 11 months ago
- ☆53Updated last year
- Understand and test language model architectures on synthetic tasks.☆217Updated 2 weeks ago
- ☆51Updated 7 months ago
- Triton Implementation of HyperAttention Algorithm☆48Updated last year
- Flash-Muon: An Efficient Implementation of Muon Optimizer☆128Updated last week
- Code for exploring Based models from "Simple linear attention language models balance the recall-throughput tradeoff"☆235Updated 2 weeks ago
- Stick-breaking attention☆57Updated this week
- ☆147Updated 2 years ago
- Layer-Condensed KV cache w/ 10 times larger batch size, fewer params and less computation. Dramatic speed up with better task performance…☆149Updated 2 months ago
- Token Omission Via Attention☆127Updated 8 months ago
- ☆76Updated 3 months ago
- ☆79Updated 10 months ago
- Load compute kernels from the Hub☆172Updated this week
- Here we will test various linear attention designs.☆59Updated last year
- This repo is based on https://github.com/jiaweizzhao/GaLore☆28Updated 9 months ago
- ☆51Updated last year