VORTA: Efficient Video Diffusion via Routing Sparse Attention

TL;DR - VORTA accelerates video diffusion transformers by sparse attention and dynamic routing, achieving speedup with negligible quality loss.

Quick Start

1. Download the checkpoints into the ./results directory under the VORTA GitHub code repository.

git lfs install
git clone [email protected]:anonymous728/VORTA
# mv VORTA/<model_name> results/, <model_name>: wan-14B, hunyuan; e.g.
mv VORTA/wan-14B results/

Other alternative methods to download the models can be found here.

2. Follow the README.md instructions to run the sampling with speedup. 🤗