facebookresearch/mae

the provided pretrained weight only includes the encoder weight, not includes the decoder weight

wangqingbin opened this issue · 2 comments

It seems that the provided pretrained weight only includes the encoder weight, not includes the decoder weight, could you provide the entire weight for pretraining model. Thank you very much!

See https://dl.fbaipublicfiles.com/mae/pretrain/mae_pretrain_vit_base_full.pth. The other two files mae_pretrain_vit_large_full.pth and mae_pretrain_vit_huge_full.pth are in the same dir too. Note these models reconstruct normalized pixels.

Could you confirm the hyperparameters used for training the model linked above? I need to train them for a few more epochs so as to compare with a related method I'm working on. Thanks so much!