What are you trying to evaluate or what issue are you running into? Are you comparing it with another model or seeing unexpected behavior?