search

Alpha-VL / ConvMAE

ConvMAE: Masked Convolution Meets Masked Autoencoders
MIT License
477 stars 41 forks source link

Question about ConvMAE-v2 #19

Closed z-jiaming closed 1 year ago

z-jiaming commented 1 year ago

Thank you for your excellent work!

When I load ConvMAE-v2-Base pretrained checkpoints [https://drive.google.com/file/d/1gykVKNDlRn8eiuXk5bUj1PbSnHXFzLnI/view?usp=sharing], it has cls_token parameter, which not in models_convmae.py.

Does ConvMAE-v2 model different from models_convmae.py in some details, thanks!

gaopengpjlab commented 1 year ago

Compared with ConvMAE-V1, ConvMAE-V2 add an extra CLS token. We will share paper and code soon.

z-jiaming commented 1 year ago

Thanks for your reply! But I'm confused that if models_convmae.py and models_convvit.py is ConvMAE-v1, why your released pretrained checkpoints for ConvMAE_Base not have mask_token, stage1_output and stage2_output, but ConvMAE-v2 provided? Did you filter them because models_convvit dosen't need them?

Thanks!!

gaopengpjlab commented 1 year ago

We filter out mask_token, stage1_output and stage2_output from the ConvMAE Base checkpoint. ConvMAE-V2 do not need mask_token, stage1_output and stage2_output.

z-jiaming commented 1 year ago

Thanks a lot ! Looking forward to ConvMAE-v2 !