Topics / #flash-attention

Flash-attention projects on GitHub

Memory-efficient attention kernels accelerating transformer training.

As of October 17, 2026, the most-starred Flash-attention projects on GitHub are Qwen (21.9K stars), mamba (18.9K stars) and HRM (12.7K stars). GitHub.Locker tracks 11 Flash-attention repositories with 84.6K stars combined. The fastest-growing on October 16 was LTX-2, with +12 stars in a day.

11Projects

84.6KStars in total

+44Stars in a day

Browse in the catalog →github.com/topics/flash-attention ↗

Hot right now

Most stars gained in a day · Oct 16

+12

LTX-2

Lightricks/LTX-2 · ★ 9.8K

The official Python package for LTX-2, Lightricks' audio-video foundation model: one DiT-based model that generates video with synchronized audio from text or an image, in several quality/speed modes.

lectures

gpu-mode/lectures · ★ 6.9K

This is the supplementary material repository for the GPU MODE lectures YouTube channel — a growing collection of notebooks, slides and code from each numbered lecture on GPU and CUDA programming.

Qwen

QwenLM/Qwen · ★ 21.9K

The official repo of Qwen (通义千问, Tongyi Qianwen) — Alibaba Cloud's family of chat and pretrained large language models, released in sizes from 1.8B up to 72B parameters on Hugging Face and ModelScope.

LeetCUDA

xlite-dev/LeetCUDA · ★ 12.1K

LeetCUDA is an open-source book of CUDA learning notes for beginners, paired with a library of 200+ hand-written CUDA kernels (elementwise ops, HGEMM, Flash-Attention, etc.) and 100+ CUDA/LLM blog posts.

The all-time top

11 projects · by stars

#1 · QwenLM/Qwen
Qwen
The official repo of Qwen (通义千问, Tongyi Qianwen) — Alibaba Cloud's family of chat and pretrained large language models, released in sizes from 1.8B up to 72B parameters on Hugging Face and ModelScope.
21.9K
#2 · state-spaces/mamba
mamba
Mamba is a state space model architecture for sequence modeling with linear-time scaling. It shows promising performance on information-dense data like language modeling, where earlier subquadratic models fell short of Transformers; this repo is its PyTorch implementation.
18.9K
#3 · sapientinc/HRM
HRM
The official release of the Hierarchical Reasoning Model (HRM): a brain-inspired recurrent architecture that reaches significant computational depth with only 27 million parameters, trained on ~1000 examples without pretraining or chain-of-thought data.
12.7K

4
LeetCUDA
12.1Kxlite-dev/LeetCUDA · Cuda
LeetCUDA is an open-source book of CUDA learning notes for beginners, paired with a library of 200+ hand-written CUDA kernels (elementwise ops, HGEMM, Flash-Attention, etc.) and 100+ CUDA/LLM blog posts.
5
LTX-2
9.8KLightricks/LTX-2 · Python
The official Python package for LTX-2, Lightricks' audio-video foundation model: one DiT-based model that generates video with synchronized audio from text or an image, in several quality/speed modes.
6
lectures
6.9Kgpu-mode/lectures · Jupyter Notebook
This is the supplementary material repository for the GPU MODE lectures YouTube channel — a growing collection of notebooks, slides and code from each numbered lecture on GPU and CUDA programming.
7
ai-system-design
672amitshekhariitbhu/ai-system-design · Markdown
A free, open-source guide that teaches AI System Design step by step: how to combine GPUs, inference servers, caches, vector databases, AI agents, guardrails, and evals into one system that is fast, cheap, reliable, and safe.
8
vdn-minimax-h3
617OpenVDN/vdn-minimax-h3 · Python
VDN-Minimax-H3 is a hybrid-attention video generation model built on MiniMax H3 that renders video faster than it plays, supporting text-to-video-audio, image-to-video-audio, first/last-frame-to-video and more from one checkpoint.
9
GLiClass
563Knowledgator/GLiClass · Python
GLiClass is an efficient zero-shot model for sequence classification, inspired by GLiNER: give it a text and a list of candidate labels, and it classifies the text with accuracy comparable to cross-encoders.
10
EasyWAM
401OpenMOSS/EasyWAM · Python
EasyWAM is a unified research codebase for training, fine-tuning, and evaluating World Action Models (WAMs) — models that jointly predict future video and robot actions for embodied-AI and robotics research.
11
hyperflow
146Video-Rebirth/hyperflow · Python
An 8-step LoRA adapter for MiniMax-H3, a video+audio generation model, trained by data-free flow self-distillation. This repository holds the loader and example scripts that run it on the diffusers Modular Pipeline.