[x] unfoldNd.Unfold2d should behave identically to torch.nn.Unfold
[x] unfoldNd.Unfold1d should behave identically to adding a dummy dimension to the input, then using torch.nn.Unfold, then squeezing out the dummy dimension
[x] The unfold operation's output can be used to perform convolution by matrix-multiplication with a matrix-view of the kernel. The result of such a convolution performed with the output of unfoldNd.unfold{1,2,3}d must be identical to the convolution of torch.nn.Conv{1,2,3}d
Potential test settings are:
unfoldNd.Unfold2d
should behave identically totorch.nn.Unfold
unfoldNd.Unfold1d
should behave identically to adding a dummy dimension to the input, then usingtorch.nn.Unfold
, then squeezing out the dummy dimensionunfoldNd.unfold{1,2,3}d
must be identical to the convolution oftorch.nn.Conv{1,2,3}d