I want to know the how to calculate deconvolution layer with cudnn.
I’m looking for an example that implements something, but I could not find it.
For example, Caffe doesn’t have the implementation of deconvolution layer with cudnn.
Is it because of cudnn has no function to calculate deconvolution layer?
Or is there a way to calculate deconvolution layer?
By the way, I have an idea to calculate convolution layer.
It is the easy idea that relocation a tensor memory and invoke convolution method of cudnn.
But I think that it is not efficiently.