Read the FAQs and search the forum before posting a new topic.
This forum is for reporting errors with the Training process. If you want to get tips, or better understand the Training process, then you should look in the Training Discussion forum.
Please mark any answers that fixed your problems so others can find the solutions.
Using Phaze-A/StoJo with efficient_v2_s and after 14000 iterations the preview is still just showing a mask (plain white where the face should be, surrounded by pink mask).
Is this normal? Are there certain mask types I should be using for Phaze-A?
I have been using 3 maskers: Bisect-Fp + Vgg-Clear + VggObstructed.
Using the same images with Villian the preview works fine after a few iterations.
Thanks
Most likely you will need to delete your model, lower the learning rate (try 3.5e-5, if that doesn't work, then 1e-5) and start again.
Tried with re-extracting with some other masks and making sure the trainer loss setting mask matched... had some luck getting preview to work with UnetDlf (or no mask), but that isn't actually the mask I want to use.
Using Bisenet-Fp only now but still have the preview issue where I only see a mask.
Hey, its been a long time, but I have just reestarted training on faceswap, now that I have time. I wish someday I will be able of doing realistic swaps.
I tryed using Phaze-A, and had to put the efficentnet_v2_s with stoho, like you did, and I run in to the same problem. Its interesting, is looks like it cant read the pixels.
This is how it looks like with 0 iterations
After a while it turns yellow instead of blue
Most likely you will need to delete your model, lower the learning rate (try 3.5e-5, if that doesn't work, then 1e-5) and start again.
My word is final
