TensorFlow optimization code GAN, resume after first epoch dont go

Hi, i need review and optimization for mine GAN code,in particular I need to resume training without going back to the first epoch, so far the tests done have not brought results, who can help me and explain as well as correct?