| import os | |
| import torch | |
| MODEL_PATH = 'pretrained_models' | |
| pt_lightnings = [ | |
| 'vqgan_imagenet_f16_1024/ckpts/last.ckpt', | |
| 'vqgan_imagenet_f16_16384/ckpts/last.ckpt', | |
| 'vq-f8-n256/model.ckpt', | |
| 'vq-f8/model.ckpt', | |
| ] | |
| pts = [ | |
| 'vqgan_imagenet_f16_1024/ckpts/last.pth', | |
| 'vqgan_imagenet_f16_16384/ckpts/last.pth', | |
| 'vq-f8-n256/model.pth', | |
| 'vq-f8/model.pth', | |
| ] | |
| for pt_l, pt in zip(pt_lightnings, pts): | |
| pt_l_weight = torch.load(os.path.join(MODEL_PATH, pt_l), map_location='cpu') | |
| pt_weight = { | |
| 'state_dict': pt_l_weight['state_dict'] | |
| } | |
| pt_path = os.path.join(MODEL_PATH, pt) | |
| torch.save(pt_weight, pt_path) | |
| print(f'saving to {pt_path}') | |