Ok, the probably simplest way is file_copy() in any INITRUN to copy the Debug.csv from the previous training loop to a target file name. Or use an individual neural() function that stores the signals.