Torch Architecture in Pytorch


Is it possible to implement the following architecture in pytorch?

I couldn’t find SpatialConvolutionMM,SpatialCrossMapLRN and nn.Normalize


Any help here?
Just point me to the right docs

CrossMapLRN is missing and isn’t wrapped yet. ConvolutionMM is Conv2d and nn.Normalize is simply just implementable using torch.* operations.

Hi Smth, thanks for your reply.
But OpenFace has inception layer, any such layer in Pytorch?

Hi Lee,
Have you worked this out? I meet the same situation these days, maybe we can discuss how to implement this in pytorch.

yifan, google for inception pytorch, you will see some implementations.