vqnhat

Thank you for your reply. The face images have been aligned by retina face before training. The pretrained model was loaded directly by torch.load()