pub fn tensor_to_array<const N: usize, B: Backend>(
    x: Tensor<B, 1>
) -> [usize; N]