Advice regarding training DCGAN

Would love to share your experience on the ideal experimental setup for training a DCGAN for generating faces.

We have two datasets:

  • LFW (13k images)
  • CelebA (200k images)

Starting from here. Planning to use a V100.