aiqm / aimnet

Atoms In Molecules Neural Network Potential

AIMNet - Atoms In Molecules Neural Network Potential

This repository contains reference AIMNet implementation along with some examples and menchmarks

Requirements

• pytorch-1.1
• numpy
• pyyaml
• ase
• nglview (for demos)

Citation

Accurate and transferable multitask prediction of chemical properties with an atoms-in-molecules neural network

Roman Zubatyuk, Justin S. Smith, Jerzy Leszczynski, Olexandr Isayev Science Advances 2019: Vol. 5, no. 8, eaav6490 DOI: 10.1126/sciadv.aav6490

aimnet's Issues

Installation problem

The package could not be installed!

Retraining model with D4 correction

Hi,

I am interested in retraining the CHONSFCl-SMD model with the d4 correction baked into the model because GPUs are able to out compute the dispersion correction on a CPU for molecules with ~100 atoms.

Do you have plans to release the full dataset with CHONSFCl dataset? Would you be interested in collaborating with me? This model is perfect for molecular design which is what I am using it for.

Thanks,
Adam Moyer

Example with atomic charges?

Is there an example or a method to get the atomic charges? It's not obvious how to get the electrostatic components from the model.

Thanks!

Installing?

Hi,

I'm attempting to install this package for my lab, but the install instructions are non-existent. I'm able to see that the package can be found on pip by google search, but that doesn't seem to be working. So which python version is being expected, and how to I go about installing this package?

Thanks
Akshar