PaiNN-model introduction

This is a simple implementation of PaiNN model and active learning workflow for fitting interatomic potentials.
The learned features or gradients in the model are used for active learning. Several selection methods are implemented.
All the active learning codes are to be tested.

Documentation

No documentation yet.

Quick Start

How to install

This code is only tested on Python>=3.8.0 and PyTorch>=1.10.
Requirements: PyTorch Scatter(if you want to use active learning), toml, myqueue(if you want to submit jobs automatically).

$ conda install pytorch-scatter -c pyg
$ conda install -c conda-forge toml
$ python3 -m pip install myqueue
$ conda install pytorch torchvision torchaudio pytorch-cuda=11.6 -c pytorch -c nvidia
$ git clone https://github.com/Yangxinsix/PaiNN-model.git
$ cd PaiNN-model
$ python -m pip install -U .

How to use

• See train.py in scripts for training, and md_run.py for running MD simulations by using ASE.
• See al_select.py for active learning.
• See flow.py for distributing and submitting active learning jobs.