Training, validation, and testing datasets for the publication "Discrepancies and the error evaluation metrics for machine learning interatomic potentials"

Each directory contains:

1. The compressed structural data (in VASP POSCAR format with .zip file)
2. The corresponding energies, forces, distances to nearest migrating atoms, and distances to nearest vacancies data (DATA.json)
3. (Available only for testing datasets) the compressed structural data for vacancies (in VASP POSCAR format with .zip file)

DATA.json architecture:
|
|-- Forces
| |-- DFT_K4
| |-- DFT_K2 (testing sets only)
| |-- DFT_K1 (testing sets only)
|
|-- Energies
| |-- DFT_K4
| |-- DFT_K2 (testing sets only)
|
|-- Structures: [index].POSCAR.vasp
|
|-- r: distances to migrating atoms (testing sets only)
|
|-- r_v: distances to vacancies (testing sets only)
|
|-- Vacancies: [index]_vacancies.POSCAR.vasp
  (identified vacancy sites)
  (testing sets only)

If you use the datasets extensively, you may want to cite the following publication:
Y. Liu, X. He, Y. Mo, Discrepancies and error evaluation metrics for machine learning interatomic potentials, Npj Comput. Mater. 9 (2023) 174.
https://doi.org/10.1038/s41524-023-01123-3.