pub fn unfold4d<B>( x: Tensor<B, 4>, kernel_size: [usize; 2], options: UnfoldOptions, ) -> Tensor<B, 3>where B: Backend,
Applies a 4D to 3D unfold.