https://github.com/keisukeyamashita/hdnnp
Python Implementation of High Dimentional Neural Network Potential 🐍🚀✨
https://github.com/keisukeyamashita/hdnnp
hdnnp neural-networks numerical-analysis
Last synced: 7 months ago
JSON representation
Python Implementation of High Dimentional Neural Network Potential 🐍🚀✨
- Host: GitHub
- URL: https://github.com/keisukeyamashita/hdnnp
- Owner: KeisukeYamashita
- License: mit
- Created: 2018-11-27T10:32:50.000Z (almost 7 years ago)
- Default Branch: master
- Last Pushed: 2018-12-03T16:52:28.000Z (almost 7 years ago)
- Last Synced: 2024-10-11T12:46:14.225Z (about 1 year ago)
- Topics: hdnnp, neural-networks, numerical-analysis
- Language: Python
- Size: 99.6 KB
- Stars: 2
- Watchers: 2
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# High Dimensional Neural Network Potential(HDNNP)
[](https://github.com/ellerbrock/open-source-badges/)
[](https://www.python.org/downloads/release/python-330/)
[](https://opensource.org/licenses/MIT)
[](https://github.com/dwyl/esta/issues)
> This is a implementation of High Dimensional Neural Network Potential(HDNNP) designed to reproduce Density Function Theory(DFT) calculation *effectively* with high *flexibility*, *reactivity*. Based on ogura-edu/HDNNP.
**Table of Contents**
- [Install](#install)
- [Setup](#setup)
- [By Pipenv(Prefered)](#by-pipenvprefered)
- [By Anaconda](#by-anaconda)
- [Usage](#usage)
- [vasp2xyz: Convert VASP OUTCAR to xyz file](#vasp2xyz-convert-vasp-outcar-to-xyz-file)
- [merge-xyz: Merge multi xyz files into one](#merge-xyz-merge-multi-xyz-files-into-one)
- [Maintainer](#maintainer)
- [Reference](#reference)
## Install
Install this project by `git`.
```shell
$ git clone https://github.com/ogura-edu/HDNNP.git
# or if using ssh
$ git clone git@github.com:ogura-edu/HDNNP.git
```
This project uses [Pipenv](https://github.com/pypa/pipenv) for development workflow. If you don't have it, run this command to install.
**macOS**
```shell
$ brew install pipenv
```
**other**
```shell
# please run after installing python
$ pip install pipenv
```
## Setup
### By Pipenv(Prefered)
Same as by anaconda, but you need to install python rather than installing anaconda.
This bug will be fixed in near future release(ref: [pythonfinder + pyenv + anaconda issue](https://github.com/pypa/pipenv/issues/3044)).
Set environmental variable `PIPENV_VENV_IN_PROJECT` to `1` to create your VM into this project dir(`/path/to/HDNNP/.venv`).
```shell
export PIPENV_VENV_IN_PROJECT = 1
```
For macOS users, you need to install `mpich` before installing dependencies.
```shell
# Only for macOS users.
#
# NOTE: Installing both mpich and openmpi will conflict
#
$ brew install mpich
# or
$ brew install openmpi
```
Setup your enviroments.
```shell
# Install dependencies
$ pipenv install
# activate your VM
$ pipenv shell
# For example...
(HDNNP) $ hdnnpy training
# deactivate
(HDNNP) $ exit
```
### By Anaconda
Using anaconda is prefered because it is basically faster than Pipenv.
Install anaconda and activate your VM.
```shell
$ ANACONDA_VERSION = [YOUR_ANACODA_VERSION]
$ pyenv install $ANACONDA_VERSION
$ pyenv local $ANACONDA_VERSION
$ conda env create -n HDNNP --file condaenv.yaml
$ echo ". ${HOME}/.pyenv/versions//etc/profile.d/conda.sh" > ~/.bashrc
# activate
$ conda activate HDNNP
# install this program using pip
(HDNNP) $ pip install --editable .
# For example...
(HDNNP) $ hdnnpy training
# deactivate
(HDNNP) $ conda deactivate
```
**NOTE**
There is no
- ChainerMN
- Chainer v5
on the Anaconda Cloud, so you still have to install these packages by `pip`.
And these is a bug that if you install anaconda by `pyenv`, `pipenv` will fail to start(ref: [pythonfinder + pyenv + anaconda issue](https://github.com/pypa/pipenv/issues/3044)).
## Usage
### vasp2xyz: Convert VASP OUTCAR to xyz file
Use `vasp2xyz` command to convert OUTCAR to xyz file.
```shell
hdnnpy vasp2xyz [PREFIX] [PATH_TO_OUTCAR] [OUTPUT_FILE]
```
### merge-xyz: Merge multi xyz files into one
```shell
hdnnpy merge-xyz [STEPS] [PATH_TO_XYZ_FILES_DIR] [OUTPUT_FILE]
```
## Maintainer
- KeisukeYamashita: maintainer and creator
## Reference
- Jörg Behler. First Principle Neural Network Potentials for Reactive Simulations of Large Molecular and Condensed System, 2007