Yes, the manual was misleading. "Training run" was referring to a loop cycle. I'll make the description more clear. You can also use the "model" parameter as an index of a particular model, and store them in R in a list of lists, f.i. Models[[model+1]].