mindspore.mint.index_select

mindspore.mint.index_select(input, dim, index)

Generates a new Tensor that accesses the values of input along the specified dim dimension using the indices specified in index. The new Tensor has the same number of dimensions as input, with the size of the dim dimension being equal to the length of index, and the size of all other dimensions will be unchanged from the original input Tensor.

Note

The value of index must be in the range of [0, input.shape[dim]), the result is undefined out of range.

Parameters

• input (Tensor) – The input Tensor.

• dim (int) – The dimension to be indexed.

• index (Tensor) – A 1-D Tensor with the indices.

Returns

Tensor, has the same dtype as input Tensor.

Raises

• TypeError – If input or index is not a Tensor.

• TypeError – If dim is not int number.

• ValueError – If the value of dim is out the range of [-input.ndim, input.ndim - 1].

• ValueError – If the dimension of index is not equal to 1.

Supported Platforms:

Ascend

Examples

>>> import mindspore
>>> from mindspore import Tensor, mint
>>> import numpy as np
>>> input = Tensor(np.arange(16).astype(np.float32).reshape(2, 2, 4))
>>> print(input)
[[[ 0.  1.  2.  3.]
[ 4.  5.  6.  7.]]
[[ 8.  9. 10. 11.]
[12. 13. 14. 15.]]]
>>> index = Tensor([0,], mindspore.int32)
>>> y = mint.index_select(input, 1, index)
>>> print(y)
[[[ 0.  1.  2.  3.]]
[[ 8.  9. 10. 11.]]]