Depthwise-Separable-Convolution_Pytorch

Depthwise-Separable-Convolution_Pytorch copied to clipboard

Hi, thanks a lot for your code. Trying to implement depthwise separable convolutions into my network to make it more computational efficient, I am asking myself, how to set the parameter kernels_per_layer properly. Do you have any recommendations what has to be considered to select it wisely? Many thanks!