Accurate_ratio not improving with DPGEN

[lbyqtttt]

Hi, dear all
I encountered an issue when using dpgen to generate datasets. After completing one iteration, I found that the accurate_ratio actually decreased. Below are the accurate_ratios for iter0 and iter1, respectively.

Additionally, when I used dptest to evaluate the accuracy of the potential functions from iter0 and iter1 on the iter0.02.fp dataset, I observed almost no improvement. Furthermore, both results are twice as high as the accuracy obtained by testing the initial training set with dptest.


and this is the initial dataset tested by dptest

Could anyone suggest why the accuracy might be decreasing after iterations and why dptest results show little improvement? Any suggestions on how to address this would be greatly appreciated. Thank you!