iags
    tensor([[0, 1, 2],[3, 4, 5], [6, 7, 8])
    >>> s = torch.sparse.spdiags(diags, torch.tensor([0, -1, -2]), (3, 3), layout=torch.sparse_csr)
    >>> s
    tensor(crow_indices=tensor([0, 1, 3, 6]),
           col_indices=tensor([0, 0, 1, 0, 1, 2]),
           values=tensor([0, 3, 1, 6, 4, 2]), size=(3, 3), nnz=6,
           layout=torch.sparse_csr)
    >>> s.to_dense()
    tensor([[0, 0, 0],
            [3, 1, 0],
            [6, 4, 2]])

Set partial diagonals of a large output::

    >>> diags = torch.tensor([[1, 2], [3, 4]])
    >>> offsets = torch.tensor([0, -1])
    >>> torch.sparse.spdiags(diags, offsets, (5, 5)).to_dense()
    tensor([[1, 0, 0, 0, 0],
            [3, 2, 0, 0, 0],
            [0, 4, 0, 0, 0],
            [0, 0, 0, 0, 0],
            [0, 0, 0, 0, 0]])

.. note::

    When setting the values along a given diagonal the index into the diagonal
    and the index into the row of :attr:`diagonals` is taken as the
    column index in the output. This has the effect that when setting a diagonal
    with a positive offset `k` the first value along that diagonal will be
    the value in position `k` of the row of :attr:`diagonals`

Specifying a positive offset::

    >>> diags = torch.tensor([[1, 2, 3], [1, 2, 3], [1, 2, 3]])
    >>> torch.sparse.spdiags(diags, torch.tensor([0, 1, 2]), (5, 5)).to_dense()
    tensor([[1, 2, 3, 0, 0],
            [0, 2, 3, 0, 0],
            [0, 0, 3, 0, 0],
            [0, 0, 0, 0, 0],
            [0, 0, 0, 0, 0]])
c