Sorry to bother the author again and again, I would like to ask a few additional questions: 1. How much video memory should I use with your model for my dataset sizes of 128 and 256? 2. Do I need to increase the middle layer of the profile for larger dataset sizes?