The ability to train with small datasets with high accuracy

[poyaii]

Hello. Can I train a potential with a radius of 6 angstrom for a small structure with three different atoms, a lattice size of 8 * 7 * 7 angstrom, and 54 atoms?! Does the DeepMD package have the capability to work with small datasets and achieve acceptable accuracy? I would appreciate it if you could guide me or provide an example of a small dataset trained with this package. Thank you.

[wanghan-iapcm]

Training with small cells is valid in deepmd-kit.

[poyaii]

Thank you, wanghan. Could you give me an example of a small structure that has been trained? Does the cutoff radius have to be larger than half of the big box? Because I considered the cutoff radius larger than half of the box and my structure collapsed! ? Thank you.

[wanghan-iapcm]

no restriction on the size of cell.