When using NEP to train the MLP, there is some information missing in the input xyz file, causing the RMSE in the loss.out file to be nan. Maybe it's useful to check whether the input files missing some of the information firstly, and when this is encountered to just abort the program and report the error?
When using NEP to train the MLP, there is some information missing in the input xyz file, causing the RMSE in the loss.out file to be nan. Maybe it's useful to check whether the input files missing some of the information firstly, and when this is encountered to just abort the program and report the error?