mindspore.ops.TensorScatterSub

查看源文件
class mindspore.ops.TensorScatterSub[源代码]

根据指定的更新值 input_x 和输入索引 indices,进行减法运算更新输出Tensor的值。当同一索引有不同更新值时,更新的结果将是累积减法的结果。此操作与 mindspore.ops.ScatterNdSub 类似,只是更新后的结果是通过算子output返回,而不是直接原地更新input。 更多参考详见 mindspore.ops.tensor_scatter_sub()

\[output\left [indices \right ] = input\_x- update\]
输入:
  • input_x (Tensor) - 输入Tensor。 input_x 的维度必须大于等于indices.shape[-1]。

  • indices (Tensor) - 输入Tensor的索引,数据类型为int32或int64,rank必须大于等于2。

  • updates (Tensor) - 指定与 input_x 相减操作的Tensor,其数据类型与 input_x 相同。并且shape应等于 \(indices.shape[:-1] + input\_x.shape[indices.shape[-1]:]\)

输出:

Tensor,shape和数据类型与输入 input_x 相同。

支持平台:

Ascend GPU CPU

样例:

>>> 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)
>>> op = ops.TensorScatterSub()
>>> # 5, Perform the subtract operation for the first time:
>>> #      first_input_x = input_x[0][0] - updates[0] = [[-1.1, 0.3, 3.6], [0.4, 0.5, -3.2]]
>>> # 6, Perform the subtract operation for the second time:
>>> #      second_input_x = input_x[0][0] - updates[1] = [[-3.3, 0.3, 3.6], [0.4, 0.5, -3.2]]
>>> output = op(input_x, indices, updates)
>>> print(output)
[[-3.3000002  0.3        3.6      ]
 [ 0.4        0.5       -3.2      ]]