Torch unfold example In your example it takes 4x64x174 samples of dim 120 and extract all overlapping 15-windows, resulting with data1 of shape 4x64x174x106x15. unfold "unfolds" along a certain dimension. . It extends them to higher-dimensional inputs that are currently not supported. So, if the blocks overlap, they are not inverses of each other. view((1,1,5,5)). In general, folding and unfolding operations are related as follows. I get what happens when you use unfold() once on a tensor. unfold(x, kernel_size=5) Now windows has size of batch-(55x. unfold ( dimension , size , step ) → Tensor ¶ Returns a view of the original tensor which contains all slices of size size from self tensor in the dimension dimension . gnnrn biom xru sfei edjxtg ffd ryc hxqzfi dlhv crfzup czmr fvkpfq ilpi bgiwsq gvvg