mindspore.ops.tensor_scatter_max

mindspore.ops.tensor_scatter_max(input_x, indices, updates)[source]

By comparing the value at the position indicated by indices in input_x with the value in the updates, the value at the index will eventually be equal to the largest one to create a new tensor.

The last axis of the index is the depth of each index vector. For each index vector, there must be a corresponding value in updates. The shape of updates should be equal to the shape of input_x[indices].

\[output\left [indices \right ] = \max(input\_x, update)\]

Note

  • On GPU, if some values of the indices are out of bound, instead of raising an index error, the corresponding updates will not be updated to self tensor.

  • On CPU, if some values of the indices are out of bound, raising an index error.

  • On Ascend, out of bound checking is not supported, if some values of the indices are out of bound, unknown errors may be caused.

Parameters

  • input_x (Tensor) – The input tensor. The dimension of input_x must be no less than indices.shape[-1].

  • indices (Tensor) – The index of input tensor whose data type must be int32 or int64. The rank must be at least 2.

  • updates (Tensor) – The tensor to update the input_x tensor, has the same type as input, and updates.shape should be equal to \(indices.shape[:-1] + input\_x.shape[indices.shape[-1]:]\).

Returns

Tensor, has the same shape and type as input_x.

Raises

  • TypeError – If dtype of indices is neither int32 nor int64.

  • ValueError – If length of shape of input_x is less than the last dimension of shape of indices.

  • RuntimeError – If a value of indices is not in input_x on CPU backend.

Supported Platforms:

GPU CPU

Examples

>>> import mindspore
>>> import numpy as np
>>> from mindspore import Tensor, ops
>>> input_x = Tensor(np.array([[-0.1, 0.3, 3.6], [0.4, 0.5, -3.2]]), mindspore.float32)
>>> indices = Tensor(np.array([[0, 0], [0, 0]]), mindspore.int32)
>>> updates = Tensor(np.array([1.0, 2.2]), mindspore.float32)
>>> # Next, demonstrate the approximate operation process of this operator:
>>> # 1, indices[0] = [0, 0], indices[1] = [0, 0]
>>> # 2, And input_x[0, 0] = -0.1
>>> # 3, So input_x[indices] = [-0.1, -0.1]
>>> # 4, Satisfy the above formula: input_x[indices].shape=(2) == updates.shape=(2)
>>> output = ops.tensor_scatter_max(input_x, indices, updates)
>>> # 5, Perform the max operation for the first time:
>>> #      first_input_x = Max(input_x[0][0], updates[0]) = [[1.0, 0.3, 3.6], [0.4, 0.5, -3.2]]
>>> # 6, Perform the max operation for the second time:
>>> #      second_input_x = Max(input_x[0][0], updates[1]) = [[2.2, 0.3, 3.6], [0.4, 0.5, -3.2]]
>>> print(output)
[[ 2.2  0.3  3.6]
 [ 0.4  0.5 -3.2]]