mindspore.ops.argsort

mindspore.ops.argsort(input, axis=- 1, descending=False)[source]

Return the indices that sort the tensor along the specified axis.

Note

The Ascend backend only supports sorting the last dimension.

Parameters

input (Tensor) – The input tensor.
axis (int) – Specify the axis to sort along. Default -1 .
descending (bool) – Specify the sorting order (ascending or descending).

Returns

Tensor

Supported Platforms:: Ascend GPU CPU

Examples

>>> import mindspore
>>> # case 1: 1-dimensional sort
>>> input = mindspore.tensor([1, 3, 5, 4, 2, 1])
>>> mindspore.ops.argsort(input)
Tensor(shape=[6], dtype=Int32, value= [0, 5, 4, 1, 3, 2])
>>>
>>> # case 2: multi-dimensional sort
>>> input = mindspore.tensor([[2, 1, 3],
...                           [6, 4, 3]])
>>> mindspore.ops.argsort(input, axis=1)
Tensor(shape=[2, 3], dtype=Int32, value=
[[1, 0, 2],
 [2, 1, 0]])
>>> mindspore.ops.argsort(input, axis=1, descending=True)
Tensor(shape=[2, 3], dtype=Int32, value=
[[2, 0, 1],
 [0, 1, 2]])