NepTrain
An automated toolkit for training neuroevolution potential (NEP), integrating tools like GPUMD, VASP, and NEP for streamlined workflows including perturbation, active learning, single-point energy calculations, and potential training.
Install / Use
/learn @aboys-cb/NepTrainREADME
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
- Join the community chat (https://qm.qq.com/q/wPDQYHMhyg)
- Raise issues and engage in discussions via GitHub issues
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 slurm
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
