System requirements:
Software:
torch - pyTorch
torchANI - pyTorch implementation of ANAKIN-ME potential
ASE - Atomic Simulation Environment

Versions tested on:
torch - 2.2.3
ASE - 3.22.1

Installation guide:
conda create -name tutorial_env numpy scipy matplotlib jupyter
source activate tutorial_env
pip3 install torch torchvision torchaudio
pip install torchani
pip install --upgrade --user ase

Install time: 5 minutes

Demo:

Instructions to run on data:
Activate conda environment
run: jupyter notebook
Script should run without further modification
Expected output:
Found in Jupyter notebook
Expected run time:
30 minutes (CPU)
2 mintues (GPU)

Instructions for use:
Ensure that directory contains the following files before running:
10water.xyz
model_alnano_ef-it8-1.info
model_alnano_ef-it8-1 (provided as supplementary data) directory containing:
rHCNO-5.2R_32-3.5A_a8-4.params
sae_linfit.dat
train0 - train7 directories each containing:
rHCNO-5.2R_32-3.5A_a8-4.params
sae_linfit.dat
inputtrain.ipt
output.opt
tolerance.dat
cost.dat
networks directory containing:
36 files total
9 files for H, C, N, and O each
4 .nnf files (one for H, C, N, and O)
16 .bparam files (4 layers, 4 elements)
16 .wparam files (4 layers, 4 elements)
