mindspore.ops.tensor_split

mindspore.ops.tensor_split(input, indices_or_sections, axis=0)

Split the input tensor into multiple subtensors according to the specified indices or chunks.

Parameters

• input (Tensor) – The input tensor.

• indices_or_sections (Union[int, tuple(int), list(int)]) –

  The specified indices or chunks.

  • If it is an integer, input tensor will be split into indices_or_sections sections.

    • If $input.shape[axis]$ can be divisible by indices_or_sections, sub-sections will have equal size $input.shape[axis]/n$ .

    • If $input.shape[axis]$ can not be divisible by indices_or_sections, the first $input.shape[axis]modn$ sections will have size $input.shape[axis]//n+1$ , and the rest will have size $input.shape[axis]//n$ .

  • If it is a tuple(int) or list(int) type, it represts indices and the input tensor will be split at the indices.

• axis (int, optional) – The axis along which to split. Default 0 .

Returns

Tuple of tensors.

Supported Platforms:

Ascend GPU CPU

Examples

>>> import mindspore
>>> input = mindspore.tensor([0, 1, 2, 3, 4, 5, 6, 7])
>>> mindspore.ops.tensor_split(input, 3)
(Tensor(shape=[3], dtype=Int64, value= [0, 1, 2]),
 Tensor(shape=[3], dtype=Int64, value= [3, 4, 5]),
 Tensor(shape=[2], dtype=Int64, value= [6, 7]))
>>> input = mindspore.tensor([0, 1, 2, 3, 4, 5, 6])
>>> mindspore.ops.tensor_split(input, 3)
(Tensor(shape=[3], dtype=Int64, value= [0, 1, 2]),
 Tensor(shape=[2], dtype=Int64, value= [3, 4]),
 Tensor(shape=[2], dtype=Int64, value= [5, 6]))
>>> mindspore.ops.tensor_split(input, (1, 6))
(Tensor(shape=[1], dtype=Int64, value= [0]),
 Tensor(shape=[5], dtype=Int64, value= [1, 2, 3, 4, 5]),
 Tensor(shape=[1], dtype=Int64, value= [6]))
>>> input = mindspore.tensor([[ 0,  1,  2,  3,  4,  5,  6],
...                           [ 7,  8,  9, 10, 11, 12, 13]])
>>> mindspore.ops.tensor_split(input, 3, axis=1)
(Tensor(shape=[2, 3], dtype=Int64, value=
 [[0, 1, 2],
  [7, 8, 9]]),
 Tensor(shape=[2, 2], dtype=Int64, value=
 [[ 3,  4],
  [10, 11]]),
 Tensor(shape=[2, 2], dtype=Int64, value=
 [[ 5,  6],
  [12, 13]]))
>>> mindspore.ops.tensor_split(input, (1, 6), axis=1)
(Tensor(shape=[2, 1], dtype=Int64, value=
 [[0],
  [7]]),
 Tensor(shape=[2, 5], dtype=Int64, value=
 [[ 1,  2,  3,  4,  5],
  [ 8,  9, 10, 11, 12]]),
 Tensor(shape=[2, 1], dtype=Int64, value=
 [[ 6],
  [13]]))