Hey @adrianhumphrey111,
The checkpoint is created by the first story which finishes with it. When training the core model, all training stories are glued together by checkpoints, duplicating where necessary to create separate dialogues to learn from. Your example works perfectly as implemented!