pytorch_ShuffleNet_ShuffleNetV2
pytorch_ShuffleNet_ShuffleNetV2 copied to clipboard
The struct of shufflenet may have some errors?
The MAC is much more than paper's. And the model saved size is also too large. I think may be there are some errors in your struct of shufflenet.
@vzvzx How to calculate the MAC?
@vzvzx v1 or v2
@vzvzx do you have script to calcualte MAC? could you give me one ?
The implement of v1 is not the same as the original parper. It doesn't use DWConv at the middle of the bottleneck.