Calculating Force in FermiNet

This repository contains the codebase for the paper Interatomic force from neural network based variational quantum Monte Carlo. If you use this code in your work, please cite our paper.

Note

This package only works with the jax branch of FermiNet, and make sure you store the last checkpoint when training. You also need JAX version like 0.2.28.

Tip

A new version is available at bytedance/netbos, where many features are added, including support for different neural networks and calculation of some other physical quantities.

Installation

cd <project-dir>
pip install -e .

Note that this will not install the FermiNet package.

Example

See example/main.py

Citation

@article{qian_interatomic_2022,
  title = {Interatomic Force from Neural Network Based Variational Quantum {{Monte Carlo}}},
  author = {Qian, Yubing and Fu, Weizhong and Ren, Weiluo and Chen, Ji},
  year = {2022},
  month = oct,
  journal = {The Journal of Chemical Physics},
  volume = {157},
  number = {16},
  pages = {164104},
  issn = {0021-9606},
  doi = {10.1063/5.0112344},
  copyright = {All rights reserved}
}