Great Job! When running the code I observed that the evaluate time is much larger than the training time is there any solution?