Function scatter_nd

python/paddle/tensor/manipulation.py:4873–4911

**Scatter_nd Layer** Output is obtained by scattering the :attr:`updates` in a new tensor according to :attr:`index` . This op is similar to :code:`scatter_nd_add`, except the tensor of :attr:`shape` is zero-initialized. Correspondingly, :code:`scatter_nd(index, updates, shape)`

(
    index: Tensor, updates: Tensor, shape: ShapeLike, name: str | None = None
)

def scatter_nd(
    index: Tensor, updates: Tensor, shape: ShapeLike, name: str | None = None
) -> Tensor:
    """
    **Scatter_nd Layer**

    Output is obtained by scattering the :attr:`updates` in a new tensor according
    to :attr:`index` . This op is similar to :code:`scatter_nd_add`, except the
    tensor of :attr:`shape` is zero-initialized. Correspondingly, :code:`scatter_nd(index, updates, shape)`
    is equal to :code:`scatter_nd_add(paddle.zeros(shape, updates.dtype), index, updates)` .
    If :attr:`index` has repeated elements, then the corresponding updates are accumulated.
    Because of the numerical approximation issues, the different order of repeated elements
    in :attr:`index` may cause different results. The specific calculation method can be
    seen :code:`scatter_nd_add` . This op is the inverse of the :code:`gather_nd` op.

    Args:
        index (Tensor): The index input with ndim >= 1 and index.shape[-1] <= len(shape).
                          Its dtype should be int32 or int64 as it is used as indexes.
        updates (Tensor): The updated value of scatter_nd op. Its dtype should be float32, float64.
                            It must have the shape index.shape[:-1] + shape[index.shape[-1]:]
        shape(tuple|list|Tensor): Shape of output tensor.
        name (str|None, optional): The output Tensor name. If set None, the layer will be named automatically.

    Returns:
        output (Tensor), The output is a tensor with the same type as :attr:`updates` .

    Examples:

        .. code-block:: pycon

            >>> import paddle

            >>> index = paddle.to_tensor([[1, 1], [0, 1], [1, 3]], dtype="int64")
            >>> updates = paddle.rand(shape=[3, 9, 10], dtype='float32')
            >>> shape = [3, 5, 9, 10]

            >>> output = paddle.scatter_nd(index, updates, shape)
    """
    return scatter_nd_add(zeros(shape, updates.dtype), index, updates, name)


@param_two_alias(["x", "input"], ["axis", "dim"])