Shoufa Chen

Thanks for your interest. For Swin, we use `bottleneck=dim // 12`, which can bring similar amount of parameters compared with plain ViT.

> > Thanks for your interest. > > For Swin, we use `bottleneck=dim // 12`, which can bring similar amount of parameters compared with plain ViT. > > Thank you...

For downstream tasks, please refer https://github.com/ShoufaChen/AdaptFormer/issues/1. We will update related results for downstream tasks after finishing experiments.

Yes, you are right.

Here: https://github.com/ShoufaChen/AdaptFormer/blob/main/main_video.py#L340-L348

`mid_dim` is 64 for ViT and `dim // 12` for swin transformer. dropout is 0 and s is 0.1.

Hi, thanks for your interest and questions. (1) The motivation behind the scaling factor *s* is to balance the original frozen feature and the newly updated feature when combining them...

I am sorry that this repo does not work currently because the page structure changed.

Hi, Thanks for your interest. I just finished my exam this semester. I am preparing the code for release. Please stay tuned.

> To add on, you can get arch_parameters by model.weights (line 62 of model_search.py in rnn). It is defined with Variable. Variable is not registered to model.parameters(). > > Only...