Skip to content

An automated toolkit for training neural network potentials (NEP), integrating tools like GPUMD, VASP, and NEP for streamlined workflows including perturbation, active learning, single-point energy calculations, and potential training.

License

Notifications You must be signed in to change notification settings

aboys-cb/NepTrain

Repository files navigation

PyPI Downloads Requires Python 3.10+

Installation

You can install it via pip:

pip install NepTrain

If you want to use the latest changes from the main branch, you can install it directly from GitHub:

pip install -U git+https://github.com/aboys-cb/NepTrain

Community Support

Software Architecture

It is recommended to use Python 3.10 or higher. Older versions might cause type errors. We also recommend using GPUMD version 3.9.5 or higher.

Usage

Modify the vim ~/.NepTrain file to change the pseudopotential file path. If this file doesn't exist, simply run NepTrain init once to generate it.

Creating Training Set (Optional)

Generate a perturbation training set for structures or structure files.

For example, apply a 0.03 lattice distortion and 0.1 atomic perturbation:

NepTrain perturb ./structure/Cs16Ag8Bi8I48.vasp --num 2000 --cell 0.03 -d 0.1  
NepTrain select perturb.xyz -max 100  

1. Initialization

First, initialize NepTrain. This will create a submission script in the current directory:

NepTrain init

3. Submit Job

After modifying the submission script and job configuration, you can submit the job by running the following command on a login node:

NepTrain train job.yaml

For running the job in the background, use nohup

nohup NepTrain train job.yaml &

If the job is interrupted, there will be a restart.yaml file in the directory. To resume the job, run:

NepTrain train restart.yaml

About

An automated toolkit for training neural network potentials (NEP), integrating tools like GPUMD, VASP, and NEP for streamlined workflows including perturbation, active learning, single-point energy calculations, and potential training.

Resources

License

Stars

Watchers

Forks

Packages

No packages published