mindspore.ops.index_add

mindspore.ops.index_add(x, indices, y, axis, use_lock=True, check_index_bound=True)[source]

Add the elements of input y into input x along the given axis and indices.

Note

indices is a one-dimensional tensor, and $i n d i c e s . s h a p e [0] = y . s h a p e [a x i s]$ .
The value range of the elements in indices is $[0, x . s h a p e [a x i s] - 1]$ .

Parameters

x (Union[Parameter, Tensor]) – The input parameter or tensor.
indices (Tensor) – The specified indices.
y (Tensor) – The input tensor to add to x.
axis (int) – The specified axis.
use_lock (bool, optional) – Whether to enable a lock to protect the updating process of variable tensors. Default True .
check_index_bound (bool, optional) – Whether to check index boundary. Default True .

Returns

Tensor

Supported Platforms:: Ascend GPU CPU

Examples

>>> import mindspore
>>> x = mindspore.Parameter(mindspore.tensor([[1, 2, 3], [4, 5, 6], [7, 8, 9]], mindspore.float32),
...                         name="name_x")
>>> indices = mindspore.tensor([0, 2], mindspore.int32)
>>> y = mindspore.tensor([[0.5, 1.0], [1.0, 1.5], [2.0, 2.5]], mindspore.float32)
>>> output = mindspore.ops.index_add(x, indices, y, 1)
>>> print(output)
[[ 1.5  2.   4. ]
 [ 5.   5.   7.5]
 [ 9.   8.  11.5]]