taoyang1122/adapt-image-models

T-MSA and S-MSA

Closed this issue a year ago · 3 comments

chaolistat commented a year ago

Where can I find the codes for T-MSA and S-MSA？Or where did the patch embedding reshape take place?Thanks

taoyang1122 commented a year ago

Please refer to https://github.com/taoyang1122/adapt-image-models/blob/main/mmaction/models/backbones/vit_clip.py#L80.

chaolistat commented a year ago

thanks very much, but why the only class uses 'ResidualAttentionBlock' says 'Space only'? https://github.com/taoyang1122/adapt-image-models/blob/main/mmaction/models/backbones/vit_clip.py#L199

taoyang1122 commented a year ago

Sorry for the confusion. It is just a comment I forgot to remove when cleaning the codes. You can ignore that. I will update the codes.