Is it possible to do arbitrary dimensional convolutions in pytorch? It seems like pytorch supports arbitrary dimensional tensors but there are no operations above three dimensions.
I think this is not possible, if you don’t want to wrap the CUDA code yourself: