This is the second try. The dataset was prepared more properly this time, and a trigger word was used.
The trigger word is "courage_monster", note that it uses and underscore, not a space.
I also used clip skip of 2, since it seems like that should help.




