Source: https://seerepo.com/uses/flash-attn
Markdown source: https://seerepo.com/uses/flash-attn.md
Description: 12 public repos that SeeRepo read use flash-attn. It's used most with torch, transformers, and einops.

# Open source repos that use flash-attn

12 public repos that SeeRepo read use flash-attn. It's used most with torch, transformers, and einops.

## Used most with

- torch: 6 repos
- transformers: 5 repos
- einops: 5 repos
- matplotlib: 4 repos
- accelerate: 4 repos
- peft: 4 repos
- Rich: 3 repos
- torchvision: 3 repos

## The repos, by stars

1. [RVC-Boss/GPT-SoVITS](https://seerepo.com/r/RVC-Boss/GPT-SoVITS): Python · 64,891 stars · 0.2% signed by AI — 1 min voice data can also be used to train a good TTS model! (few shot voice cloning)
2. [mudler/LocalAI](https://seerepo.com/r/mudler/LocalAI): Go · 50,345 stars · 6% signed by AI — LocalAI is the open-source AI engine. Run any model - LLMs, vision, voice, image, video - on any hardware. No GPU required.
3. [hpcaitech/ColossalAI](https://seerepo.com/r/hpcaitech/ColossalAI): Python · 41,424 stars · 0.3% signed by AI — Making large AI models cheaper, faster and more accessible
4. [OpenBMB/MiniCPM-V](https://seerepo.com/r/OpenBMB/MiniCPM-V): Python · 27,152 stars · 1.1% signed by AI — See, hear, and interact on your local devices.
5. [Wan-Video/Wan2.2](https://seerepo.com/r/Wan-Video/Wan2.2): Python · 18,418 stars — Wan: Open and Advanced Large-Scale Video Generative Models
6. [NVIDIA/TensorRT-LLM](https://seerepo.com/r/NVIDIA/TensorRT-LLM): Python · 15,038 stars · 1.7% signed by AI — TensorRT LLM provides users with an easy-to-use Python API to define Large Language Models (LLMs) and supports state-of-the-art optimizations to perform inference efficiently on NVIDIA GPUs. TensorRT LLM also contains components to create Python and C++ runtimes that orchestrate the inference execution in a performant way.
7. [ace-step/ACE-Step-1.5](https://seerepo.com/r/ace-step/ACE-Step-1.5): Python · 14,532 stars · 35.8% signed by AI — The most powerful local music generation model that outperforms almost all commercial alternatives, supporting Mac, AMD, Intel, and CUDA devices.
8. [FlagOpen/FlagEmbedding](https://seerepo.com/r/FlagOpen/FlagEmbedding): Python · 12,314 stars — Retrieval and Retrieval-augmented LLMs
9. [OpenRLHF/OpenRLHF](https://seerepo.com/r/OpenRLHF/OpenRLHF): Python · 10,152 stars · 1.8% signed by AI — An Easy-to-use, Scalable and High-performance Agentic RL Framework based on Ray (PPO & DAPO & REINFORCE++ & VLM & TIS & vLLM & Ray & Async RL)
10. [meituan-longcat/LongCat-Video](https://seerepo.com/r/meituan-longcat/LongCat-Video): Python · 9,453 stars
11. [X-PLUG/MobileAgent](https://seerepo.com/r/X-PLUG/MobileAgent): HTML · 9,379 stars —  Mobile-Agent: The Powerful GUI Agent Family
12. [tile-ai/tilelang](https://seerepo.com/r/tile-ai/tilelang): Python · 8,976 stars · 4.5% signed by AI —  Domain-specific language designed to streamline the development of high-performance GPU/CPU/Accelerators kernels

## Frequently asked questions

### What is the most popular open source repo that uses flash-attn?

RVC-Boss/GPT-SoVITS, with 64,891 stars on GitHub.

The proof: [RVC-Boss/GPT-SoVITS on GitHub](https://github.com/RVC-Boss/GPT-SoVITS)

### What other technologies is flash-attn used with?

With torch (6), transformers (5), einops (5), matplotlib (4), and accelerate (4), out of the 12 repos that use it.

### What languages are the repos that use flash-attn written in?

Python (10), Go (1), and HTML (1).

### How many repos that use flash-attn have commits signed by AI?

8 of the 12; the most, ace-step/ACE-Step-1.5, with 35.8% of its commits.

The proof: [the commits each tool signed in ace-step/ACE-Step-1.5](https://seerepo.com/r/ace-step/ACE-Step-1.5/ai)

## What is your repo built with?

Any public repo, with no account: https://seerepo.com/r/owner/repo/built, with the owner and name it has on GitHub.
