Description
  Datasets containing structures, with energies, forces and stress obtained from DFT calculations.
  Calculations where performed using VASP, see published paper for details.
  Scripts used to read data and train neural network potentials (NNPs).
  The trained NNPs for each dataset, trained using SchNetPack.

Data format
  The datasets are in the ASE database format, for details see: https://wiki.fysik.dtu.dk/ase/ase/db/db.html
  A Python script (read_db_example.py) is given as an example for how to read the data stored the ASE db format.

Units
  All quantities are given in these units: Å for coordinates, eV for energy, eV/Å for forces and eV/Å^3 for stress.

Scripts
  vasp_to_db.py
    Script for reading the data from VASP calculation into the ASE db format.

  schnet_v5_3.py and config.py
    The python script used for training the NNPs (schnet_v5_3.py) and the settings for the hyperparameters (config.py).
    For training, place the schnet_v5_3.py and config.py files in the same directory.

  read_db_example.py
    An example script for reading the data from the ASE db files in Python.

Datasets
  CA-9.db
    All 48000 images generated for the CA-9 dataset.

  CA-9_training.db
    Training images used for training the NNP_CA-9 potential.

  CA-9_validation.db
    Validation images used for training the NNP_CA-9 potential.

  RR_training.db
    The Random-Random training images used for training the NNP_RR potential, see paper for details.

  RR_validation.db
    The Random-Random validation images used for training the NNP_RR potential, see paper for details.

  BR_training.db
    The Binning-Random training images used for training the NNP_BR potential, see paper for details.

  BR_validation.db
    The Binning-Random validation images used for training the NNP_BR potential, see paper for details.

  BB_training.db
    The Binning-Binning training images used for training the NNP_BB potential, see paper for details.

  BB_validation.db
    The Binning-Binning validation images used for training the NNP_BB potential, see paper for details.

  test_images.db
    Test images used to evaluate the trained NNPs

NNPs
  The best trained potential for each dataset, NNP_CA-9, NNP_RR, NNP_BR and NNP_BB.
  These are used with SchNetPack (https://schnetpack.readthedocs.io/) and can be loaded using spk.utils.load_model()