We can run multiple models on one GPU for the router to choose, including the sleep and wakeup functionality based on the traffic monitoring status.