Optimal number of neurons in NAR network

Hi guys, I need a help with the optimal number of neurons in hidden layer of NAR network. Im using ntstool for generating the netwrok and then advanced script for any adjustments. Heres the thing I need to adjust the script in a way that the network would automatically train for a set range of neurons so say 1 to 100 neurons and save the mse for each iteration cause i dont want to do it manually like a mad man. Im sure its possible im just a noob. Heres the part of code that would need to change i guess