omertov / encoder4editing

Official implementation of "Designing an Encoder for StyleGAN Image Manipulation" (SIGGRAPH 2021) https://arxiv.org/abs/2102.02766
MIT License
945 stars 154 forks source link

Pretrained model which can be loaded with 'load_from_train_checkpoint' method #78

Closed SleyAI closed 2 years ago

SleyAI commented 2 years ago

Hello, I would like to load your pretrained model by using the "load_from_train_checkpoint" in train.py in order to continue the training. The idea is to do transfer learning by learning a few steps on new images. It looks like that the pretrained model you provided (e4e_ffhq_encode.pt) is not suitable because it's missing some information like 'discriminator_state_dict' and more. I believe it is only for inference. Do you have a model which is suitable for further training? Thank you very much!

Edit: Got it running. I just loaded the wrong model.

wileewang commented 2 years ago

Sorry to bother, do you find checkpoint of ffhq latent discriminator?