COVER's training data set is only 75 pieces. Is it training together with other training sets or training alone? If it is only training COVER, how many generations does such a small data set train?

[Xuan zhao]