Ablation study and official code problem

[xingshulicc]

First, thanks for your contribution, this paper inspired me a lot.
However, I still have some questions as follows, I hope you can answer:

1. about the code, I think the output shape of Unfold operation is not right in your code, even if it has no error
2. about the ablation study, it would be better to compare the dynamic convolution with your outlook attention. they are very similar exactly with each other, the only difference is the weights generation method. I am very interested in this.
3. according to your paper, I modified your code with my own understanding:
   https://github.com/xingshulicc/Vision-In-Transformer-Model/blob/main/outlook_attention.py.
   Hoping you can give me some advice on my code.
   Thanks again.