found reduplicated shortcut connections for the same attention module

[Karbo123]

Firstly many thanks to this great work. However, I find something weird in the codes, but I don't know whether it is intentional.

I find that SpatialAttention already has a shortcut connection in its forward function, but I also find another shortcut connection in Block's forward function. I think that the shortcuts are reduplicated twice for the same attention module, it should be better to remove the first residual connection here.

Please point out any possible mistakes in my comments, thanks.

[MenghaoGuo]

Thanks for your attention to our work.

Careful observation.

In fact, it is not a redundancy. In the experimen, we find that it can improve training stability and final result by adding a residual connection. It is also written in experiment section.

[Karbo123]

Thanks for the quick reply. I initially think it is a mistake, but now I know it is written on purpose.
However, I still think that the trick is rather elusive, and hard to understand.