mddatasetbuilder 1.1.3

A script to make molecular dynamics (MD) datasets for neural networks from given LAMMPS trajectories automatically.

MDDatasetBuilder

MDDatasetBuilder is a script to build molecular dynamics (MD) datasets for neural networks from given LAMMPS trajectories automatically.

Author: Jinzhe Zeng

Email: [e-mail]

Requirements

• Python 3.6
• Python packages: numpy, scikit-learn, ASE, GaussianRunner, tqdm, coloredlogs
• Software: Gaussian 16, DeePMD

Installation

git clone https://github.com/njzjz/mddatasetbuilder
cd mddatasetbuilder/
python3 setup.py install

You can test whether ReacNetGenerator is running normally:

python3 setup.py test

Simple example

A LAMMPS bond file and a LAMMPS dump file should be prepared.

datasetbuilder -d dump.ch4 -b bonds.reaxc.ch4_new -a C H O -n ch4 -i 25

Then you can calculate generated Gaussian files:

qmcalc -d dataset_ch4_GJf/000
qmcalc -d dataset_ch4_GJf/001

Next, prepare DeePMD datas and use DeePMD-kit to train a model.

preparedeepmd -p dataset_ch4_GJf -a C H O
cd train && dp_train train.json