Number of atoms in the deepks training system

[yycx1111]

I want to train a deepks model based on GaN system like single water molecule case and would like to know what are the requirements for the GaN system used to produce the dataset. Here are some queries to be answered, thanks!

1. What is the number of system atoms to produce the deepks training dataset in the single water molecule case？
2. What is the number of atoms of the system that should be used to do the DFT calculations to prepare the labels when preparing the deepks training dataset？
3. What are the selection criteria for the system used to prepare the deepks dataset, including number of atoms, structural requirements？