This is my first try of the software(and first post here), and i have question about "No warp" and "No flip" training that the guide wasn't clear to me on.
When you set to "No Warp" do you do this with your training data or the extracts you wish to convert(fit training)? And if it is the former, do you turn warp back on for "fit training"?
Below is where i am with training using the dfaker model at about 260,000 iterations
Input size 512
Output size 256 (This might have been a mistake)
A 11,476 images
B 21,950 images
I have kept colour augmentation on, as i believe it helps with the lighting differences between sets.
Not all the source images were added imediatly, as i initially thought i could do what i wanted with less.