Congratulations on the amazing work. I am curious about the efficiency of the method. Could you shed some light on the inference time and the memory requirement?