chengzeyi/ParaAttention
https://wavespeed.ai/ Context parallel attention that accelerates DiT model inference with dynamic caching
This project helps AI practitioners generate images and videos much faster from text or image prompts using large diffusion models like FLUX and HunyuanVideo. By intelligently parallelizing calculations and reusing past computations, it takes your prompts and delivers the final high-quality visual outputs significantly quicker. It is designed for researchers, artists, or developers working with large-scale generative AI models.
425 stars. No commits in the last 6 months.
Use this if you need to speed up the generation of high-quality images or videos using large diffusion transformer (DiT) models, especially across multiple GPUs or with dynamic caching.
Not ideal if you are working with smaller models that don't benefit from parallel processing or if your primary concern is absolute pixel-level fidelity without any acceptable trade-off for speed.
Stars
425
Forks
45
Language
Python
License
—
Category
Last pushed
Jul 05, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/transformers/chengzeyi/ParaAttention"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
lucidrains/x-transformers
A concise but complete full-attention transformer with a set of promising experimental features...
kanishkamisra/minicons
Utility for behavioral and representational analyses of Language Models
lucidrains/simple-hierarchical-transformer
Experiments around a simple idea for inducing multiple hierarchical predictive model within a GPT
lucidrains/dreamer4
Implementation of Danijar's latest iteration for his Dreamer line of work
Nicolepcx/Transformers-in-Action
This is the corresponding code for the book Transformers in Action