The reconstruction of mae_pretrain_vit_base.pth is awful, is it right?
zheng547 opened this issue · comments
Zheng commented
I use the pre-trained model from https://dl.fbaipublicfiles.com/mae/pretrain/mae_pretrain_vit_base.pth.
babar414 commented
Hi, this could be due to your pretrained model does not contained the weights of the decoder part as you may be using the pretrained model which is supposed to be used for finetuning/probing but not visualizing. You may try the model that is mentioned in the demo notebook provided by the authors at below link:
https://github.com/facebookresearch/mae/blob/main/demo/mae_visualize.ipynb