Anisotropic Analitically Modelled Density
NOTE: We are retraining the networks, to ensure reproducibility
A2MD is an ensamble of modules that aim to allow to work with electron density, model it, and predict it. It is based on the use of atom centered exponential functions with anisotropic modifications to model bonding density.
The A2MD repository contains:
- a2mdlib: read/write chemical formats.
- a2md: fit models of electron electron density
- a2mdnet: prediction of model parameters using a deep learning approach
- cdens: a fast C executable to sample density
- a2mdtests: small ensemble of molecules to test the methodology
- scripts: the methodology turned into a cli set of tools
- examples
Read the docs folder to learn to use the library and the CLI tools.
Some of the most important dependencies are:
- numpy
- torch
- torchani (1.2)
We provide a "requirements.txt" to ease installation through pip.
There is a lot of things to polish, and we'd love feedback. Don't hesitate to contact us:
bruno.czvuria@upm.es
This project is distributed under a GNU GPL v3.0, so the only restriction is to distribute closed source versions.