About the normalization

Hi, thank you for making this implementation publicly available firstly. I wonder how to initialize the statistics (normalizarion_means, normalizarion_stds) in **data.py**, and how to obtain them by my own data? Thanks! 