Lnaden / mdi-ani-driver

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Molecular Dynamics with TorchANI

This repository contains a prototype implementation of a molecular dynamics simulation using the ANI2x neural network potential. The code uses TorchANI to calculate forces, and LAMMPS to perform the simulation. Communication between the codes is controlled with the MolSSI Driver Interface (MDI).

Please note that this is a prototype and may still have some problems or improvements to be made. It's a very simple implementation based off of the principles of MDI and an introductory tutorial for TorchANI and has yet to be tested and validated.

To run this code, you should install Docker and MDIMechanic. You can find instructions for installing Docker here.

To install MDI Mechanic, make a Python environment, and do:

pip install mdimechanic

To run the code, first build in the repository using mdimechanic:

mdimechanic build

Run a minimization with

mdimechanic run --name ani-min

Then, run an MD simulation using

mdimechanic run --name ani-md

The output files will be in the lammps directory.

About


Languages

Language:Python 59.4%Language:Jupyter Notebook 36.8%Language:Dockerfile 3.8%