Experiments and Evaluations for SAIUnit

This repository contains code and experiments for evaluating the SAIUnit framework. The experiments focus on training and visualizing neural network models.

Installation

To install the required packages, run the following command:

pip install -r requirements.txt

fig3-diffrax and fig4-pde-simulation

For the Fig3-diffrax and fig4-pde-simulation experiment, we should first install the unit-aware diffrax first, please run the following command:

pip install git+https://github.com/chaoming0625/diffrax.git

python xxxx.py

Citation

If you use this code in your research, please consider citing the following paper:

@article{Wang2025IntegratingPU,
  title={Integrating physical units into high-performance AI-driven scientific computing},
  author={Chaoming Wang and Sichao He and Shouwei Luo and Yuxiang Huan and Si Wu},
  journal={Nature Communications},
  year={2025},
  volume={16},
  url={https://api.semanticscholar.org/CorpusID:277853734}
}