I would like to ask some questions about a new iteration process in DP-GEN

[twoslucky]

A new iterative process for DP-GEN is to first read the data from the training set, generate a model, test it, determine accuracy, and then perform DFT training. Is that right?If I achieve 100% accuracy in one iteration, the potential function meets the requirements. However, later on, because I wanted this potential function to describe other structures, I added a new dataset. And the accuracy of the previous iteration reached 100%. If I add a new dataset, will the machine still learn? Or is it that since the accuracy reached 100% in the previous round, the machine will no longer learn, and at this point, no matter what kind of dataset is added, it will not learn again?