Accelerating Diffusion LLM Inference via Local Determinism Propagation

🚀 Quick Start

Installation

conda create -n localleap python=3.10
conda activate localleap
conda install pytorch==2.5.1 torchvision==0.20.1 torchaudio==2.5.1 pytorch-cuda=12.4 -c pytorch -c nvidia
pip install -r requirements.txt

Inference

LLaDA:

bash scripts/llada/run.sh

Dream:

bash scripts/dream/run.sh

🙇 Acknowledgments

The codebase of LocalLeap is adapted from Fast-dLLM. We are also grateful for the following projects:

• LLaDA
• Dream