Query about Persistent Use of Masking during Inference and Finetuning in MAE about mae HOT 6 CLOSED

Hi, in finetuning, the encoder is reloaded to ViT_Classifier where the masking is discarded.

from mae.

Yes, that is the same with the original MAE, although their implementation may be slightly different.

from mae.

Thank you for your reply!
Is this way same with the original MAE?

from mae.

Thanks a lot for your kindness.
However, I can not find where the masking function off.
Could you indicate where is it?

from mae.

Here we use the encoder from MAE to build the ViT_Classifier. In the __init__ and forward here you can see anything except the masking is in use.

from mae.

I really appreciate it !

from mae.