I mean if I add a parameter as output that is required (something like

training_done=True

) in the second pipeline, that could work ?