Hi, I've observed quite good results when using a basic 256x256 transform for demo training.
I am interested in training with higher resolutions, for example, using a 512x512 or 1024x1024 transform.
Is this restyle-psp model only capable of supporting inputs of 256 pixels?
I want to use my own model in place of --stylegan_weights=pretrained_models/stylegan2-ffhq-config-f.pt.
Would it work if I use a model trained according to the instructions at https://github.com/rosinality/stylegan2-pytorch#convert-weight-from-official-checkpoints?
What kind of dataset structure is recommended for training with rosinality's StyleGAN2-PyTorch?
Thank you!!
Hi, I've observed quite good results when using a basic 256x256 transform for demo training.
I am interested in training with higher resolutions, for example, using a 512x512 or 1024x1024 transform.
Is this restyle-psp model only capable of supporting inputs of 256 pixels?
I want to use my own model in place of --stylegan_weights=pretrained_models/stylegan2-ffhq-config-f.pt.
Would it work if I use a model trained according to the instructions at https://github.com/rosinality/stylegan2-pytorch#convert-weight-from-official-checkpoints?
What kind of dataset structure is recommended for training with rosinality's StyleGAN2-PyTorch?
Thank you!!