Infinite Potential Edges

[utterwqlnut]

Infinite Potential Edges

This is a PR for adding support for infinite potential edges from https://arxiv.org/pdf/2306.10045.pdf
Beware that this does involve cython code, which means in order to use this added support, the cython code has to be compiled or there will be errors.

General

• helpers.py many changes here
  • calculate_inf_potentials method which calculates the pauli, LD, and Coulomb potentials.
  • calculate_edges_mastermethod has been updated to support infinite edge calculation, and create inf_edge_attr (weighted sum of potentials mentioned above) and inf_edge_index
  • generate_edge_features_inf method concatenates the gaussian expansion of inf_edge_attr and gaussian expansion of edge weights
• processor.py added support for above methods in helpers
• configs/example/config_potnet.yml example config for a cgcnn model using infinite edges
• matdeeplearn/preprocessor/inf_functions here there is the cython helper code for calculate_inf_potentials
  • In order to use infinite edges the cython code must be compiled
  • First setup gsl, on perlmutter you can do module load gsl, on your local device you need to set it up with the gsl-latest.tar.gz file.
  • Then run python3 setup.py build_ext --inplace