Has someone tried Global Weighted Average Pooling

Deep Learning
1

Has someone implemented this in pytorch?
I have some questions about the number of parameters in the linear transformation before the exp funcion. Can this type of pooling be implemented to be resolution independent?