Note: Running this requires a GPU and time for model inference.
Enter a model name to start actual evaluation.