mindspore.mint.index_add

mindspore.mint.index_add(input, dim, index, source, alpha=1)[source]

Accumulate the elements of alpha times source into the input by adding to the index in the order given in index. For example, if dim == 0 , index[i] == j , and alpha = -1 , then the i th row of source is subtracted from the j th row of input . The dim th dimension of source must have the same size as the length of index , and all other dimensions must match input, or an error will be raised. For a 3-D tensor, the output is defined as follows:

\begin{array}{r} \begin{array}{ll} i n p u t [i n d e x [i], :, :] + = a l p h a * s o u r c e [i, :, :] # i f d i m == 0 \\ i n p u t [:, i n d e x [i], :] + = a l p h a * s o u r c e [:, i, :] # i f d i m == 1 \\ i n p u t [:, :, i n d e x [i]] + = a l p h a * s o u r c e [:, :, i] # i f d i m == 2 \end{array} \end{array}

Warning

This is an experimental API that is subject to change or deletion.

Parameters

input (Tensor) – The input Tensor.
dim (int) – The dimension along which to index.
index (Tensor) – Add the value of "input Tensor" and source along the dimension of the dim according to the specified index value, with data type int32. The index must be 1D with the same size as the size of source in the dim dimension. The values of index should be in [0, b), where the b is the size of "input Tensor" in the dim dimension.
source (Tensor) – The input tensor with the value to add. Must have same data type as "input Tensor". The shape must be the same as "input Tensor" except the dim th dimension.

Keyword Arguments

alpha (number, optional) – The scalar multiplier for source. Default: 1.

Returns

Tensor, has the same shape and dtype as input.

Raises

TypeError – If neither index nor source is a Tensor.
ValueError – If the value of dim is out of the dimension range of source shape.
ValueError – If index rank is not the same as source rank.
ValueError – If shape of index is not 1D or size of index is not equal to dimension of source[dim].
ValueError – If the shape of source is not the same as that of input except the dim axis.

Supported Platforms:: Ascend

Examples

>>> import numpy as np
>>> import mindspore
>>> from mindspore import Tensor, mint
>>> x = Tensor(np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]]), mindspore.float32)
>>> index = Tensor(np.array([0, 2]), mindspore.int32)
>>> y = Tensor(np.array([[0.5, 1.0], [1.0, 1.5], [2.0, 2.5]]), mindspore.float32)
>>> output = mint.index_add(x, 1, index, y, alpha=1)
>>> print(output)
[[ 1.5  2.   4. ]
 [ 5.   5.   7.5]
 [ 9.   8.  11.5]]