Can someone suggest what changes should I do to the training scripts to use multiscale discriminator instead of projected discriminator?
I have tried training with projected discriminator, but the ID is not very well pasted on the target face. I have trained till 250k epochs with batch size 40.
Some tips about training can also help.
Can someone suggest what changes should I do to the training scripts to use multiscale discriminator instead of projected discriminator? I have tried training with projected discriminator, but the ID is not very well pasted on the target face. I have trained till 250k epochs with batch size 40. Some tips about training can also help.