Closed Stupid-wangnz closed 1 month ago
I didn't see any information about which pretrained model the encoder model used in the paper, maybe I overlooked it. May I ask if you can provide more information about the pretrained encoder used.
I'm using wide-resnet-50-2-export-5ae25d50.pth, it works
We used pretrained model downloaded from torchvision official website. For resnet50-2.pth, you can download it by this link.
The difference between IMAGENET1K_V1 and IMAGENET1K_V2 versions of the pretrained model lies in different training configurations.
How can I download the pretrained model mentioned in the training script. "--pretrained 2 --freeze_weight 3 --weight_name resnet50-2.pth"