.. _op_ai_onnx_Split:

Split
=====

- **Domain**: ``ai.onnx``
- **Since version**: 18

Split a tensor into a list of tensors, along the specified 'axis'.
Either input 'split' or the attribute 'num_outputs' should be specified, but not both.
If the attribute 'num_outputs' is specified, then the tensor is split into equal sized parts.
If the tensor is not evenly splittable into ``num_outputs``, the last chunk will be smaller.
If the input 'split' is specified, it indicates the sizes of each output in the split.

**Inputs**

- **input** (*T*): The tensor to split
- **split** (*tensor(int64)*): Optional length of each output. Values should be >= 0.Sum of the values must be equal to the dim value at 'axis' specified.

**Outputs**

Between 1 and ∞ outputs.

- **outputs** (*T*): One or more outputs forming list of tensors after splitting

**Attributes**

- **axis** (*int*): Which axis to split on. A negative value means counting dimensions from the back. Accepted range is [-rank, rank-1] where r = rank(input).
- **num_outputs** (*int*): Number of outputs to split parts of the tensor into. If the tensor is not evenly splittable the last chunk will be smaller.

**Type Constraints**

- **T**: Constrain input and output types to all tensor types.
  Allowed types: tensor(bfloat16), tensor(bool), tensor(complex128), tensor(complex64), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(string), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8).

Examples
--------

**test_cc_shape_inference_concat_split_even**

.. code-block:: text

    Node:
      Split(xy) -> (S1, S2)
      Attributes:
        axis = 1
        num_outputs = 2

.. code-block:: text

    Inputs:
      xy: shape=(3, 4), dtype=float32
        [[-1.        , -0.9       , -0.8       , -0.7       ],
         [-0.6       , -0.5       , -0.39999998, -0.3       ],
         [-0.19999999, -0.09999996,  0.        ,  0.10000002]]
      input_1: shape=(3, 6), dtype=float32
        [[-0.5       , -0.3       , -0.09999999,  0.10000002,  0.3       ,  0.5       ],
         [ 0.70000005,  0.9       ,  1.1       ,  1.3000001 ,  1.5       ,  1.7       ],
         [ 1.9000001 ,  2.1000001 ,  2.3       ,  2.5       ,  2.7       ,  2.9       ]]

    Outputs:
      S1: shape=(3, 10), dtype=float32
        [[0.        , 0.        , 0.10000002, 0.3       , 0.5       , 0.        ,
          0.        , 0.        , 0.        , 0.        ],
         [0.9       , 1.1       , 1.3000001 , 1.5       , 1.7       , 0.        ,
          0.        , 0.        , 0.        , 0.70000005],
         [2.1000001 , 2.3       , 2.5       , 2.7       , 2.9       , 0.        ,
          0.        , 0.        , 0.10000002, 1.9000001 ]]

**test_cc_shape_inference_concat_split_odd**

.. code-block:: text

    Node:
      Split(xy) -> (S1, S2)
      Attributes:
        axis = 1
        num_outputs = 2

.. code-block:: text

    Inputs:
      xy: shape=(3, 4), dtype=float32
        [[-1.        , -0.9       , -0.8       , -0.7       ],
         [-0.6       , -0.5       , -0.39999998, -0.3       ],
         [-0.19999999, -0.09999996,  0.        ,  0.10000002]]
      input_1: shape=(3, 6), dtype=float32
        [[-0.5       , -0.3       , -0.09999999,  0.10000002,  0.3       ,  0.5       ],
         [ 0.70000005,  0.9       ,  1.1       ,  1.3000001 ,  1.5       ,  1.7       ],
         [ 1.9000001 ,  2.1000001 ,  2.3       ,  2.5       ,  2.7       ,  2.9       ]]

    Outputs:
      S1: shape=(3, 10), dtype=float32
        [[0.        , 0.        , 0.10000002, 0.3       , 0.5       , 0.        ,
          0.        , 0.        , 0.        , 0.        ],
         [0.9       , 1.1       , 1.3000001 , 1.5       , 1.7       , 0.        ,
          0.        , 0.        , 0.        , 0.70000005],
         [2.1000001 , 2.3       , 2.5       , 2.7       , 2.9       , 0.        ,
          0.        , 0.        , 0.10000002, 1.9000001 ]]

**test_cc_split_1d_uneven_split_opset18**

.. code-block:: text

    Node:
      Split(input) -> (output_1, output_2, output_3, output_4)
      Attributes:
        num_outputs = 4

.. code-block:: text

    Inputs:
      input: shape=(7,), dtype=float32
        [1., 2., 3., 4., 5., 6., 7.]

    Outputs:
      output_1: shape=(2,), dtype=float32
        [1., 2.]
      output_2: shape=(2,), dtype=float32
        [3., 4.]
      output_3: shape=(2,), dtype=float32
        [5., 6.]
      output_4: shape=(1,), dtype=float32
        [7.]

**test_cc_split_2d_uneven_split_opset18**

.. code-block:: text

    Node:
      Split(input) -> (output_1, output_2, output_3)
      Attributes:
        axis = 1
        num_outputs = 3

.. code-block:: text

    Inputs:
      input: shape=(2, 8), dtype=float32
        [[ 1.,  2.,  3.,  4.,  5.,  6.,  7.,  8.],
         [ 9., 10., 11., 12., 13., 14., 15., 16.]]

    Outputs:
      output_1: shape=(2, 3), dtype=float32
        [[ 1.,  2.,  3.],
         [ 9., 10., 11.]]
      output_2: shape=(2, 3), dtype=float32
        [[ 4.,  5.,  6.],
         [12., 13., 14.]]
      output_3: shape=(2, 2), dtype=float32
        [[ 7.,  8.],
         [15., 16.]]

**test_cc_split_equal_parts_1d_opset18**

.. code-block:: text

    Node:
      Split(input) -> (output_1, output_2, output_3)
      Attributes:
        axis = 0
        num_outputs = 3

.. code-block:: text

    Inputs:
      input: shape=(6,), dtype=float32
        [1., 2., 3., 4., 5., 6.]

    Outputs:
      output_1: shape=(2,), dtype=float32
        [1., 2.]
      output_2: shape=(2,), dtype=float32
        [3., 4.]
      output_3: shape=(2,), dtype=float32
        [5., 6.]

**test_cc_split_equal_parts_2d**

.. code-block:: text

    Node:
      Split(input) -> (output_1, output_2)
      Attributes:
        axis = 1
        num_outputs = 2

.. code-block:: text

    Inputs:
      input: shape=(2, 6), dtype=float32
        [[ 1.,  2.,  3.,  4.,  5.,  6.],
         [ 7.,  8.,  9., 10., 11., 12.]]

    Outputs:
      output_1: shape=(2, 3), dtype=float32
        [[1., 2., 3.],
         [7., 8., 9.]]
      output_2: shape=(2, 3), dtype=float32
        [[ 4.,  5.,  6.],
         [10., 11., 12.]]

**test_cc_split_equal_parts_default_axis_opset18**

.. code-block:: text

    Node:
      Split(input) -> (output_1, output_2, output_3)
      Attributes:
        num_outputs = 3

.. code-block:: text

    Inputs:
      input: shape=(6,), dtype=float32
        [1., 2., 3., 4., 5., 6.]

    Outputs:
      output_1: shape=(2,), dtype=float32
        [1., 2.]
      output_2: shape=(2,), dtype=float32
        [3., 4.]
      output_3: shape=(2,), dtype=float32
        [5., 6.]

**test_cc_split_variable_parts_1d_opset18**

.. code-block:: text

    Node:
      Split(input, split) -> (output_1, output_2)
      Attributes:
        axis = 0

.. code-block:: text

    Inputs:
      input: shape=(6,), dtype=float32
        [1., 2., 3., 4., 5., 6.]
      split: shape=(2,), dtype=int64
        [2, 4]

    Outputs:
      output_1: shape=(2,), dtype=float32
        [1., 2.]
      output_2: shape=(4,), dtype=float32
        [3., 4., 5., 6.]

**test_cc_split_variable_parts_2d_opset18**

.. code-block:: text

    Node:
      Split(input, split) -> (output_1, output_2)
      Attributes:
        axis = 1

.. code-block:: text

    Inputs:
      input: shape=(2, 6), dtype=float32
        [[ 1.,  2.,  3.,  4.,  5.,  6.],
         [ 7.,  8.,  9., 10., 11., 12.]]
      split: shape=(2,), dtype=int64
        [2, 4]

    Outputs:
      output_1: shape=(2, 2), dtype=float32
        [[1., 2.],
         [7., 8.]]
      output_2: shape=(2, 4), dtype=float32
        [[ 3.,  4.,  5.,  6.],
         [ 9., 10., 11., 12.]]

**test_cc_split_variable_parts_default_axis_opset18**

.. code-block:: text

    Node:
      Split(input, split) -> (output_1, output_2)

.. code-block:: text

    Inputs:
      input: shape=(6,), dtype=float32
        [1., 2., 3., 4., 5., 6.]
      split: shape=(2,), dtype=int64
        [2, 4]

    Outputs:
      output_1: shape=(2,), dtype=float32
        [1., 2.]
      output_2: shape=(4,), dtype=float32
        [3., 4., 5., 6.]

**test_cc_split_zero_size_splits_opset18**

.. code-block:: text

    Node:
      Split(input, split) -> (output_1, output_2, output_3)

.. code-block:: text

    Inputs:
      input: shape=(0,), dtype=float32
        []
      split: shape=(3,), dtype=int64
        [0, 0, 0]

    Outputs:
      output_1: shape=(0,), dtype=float32
        []
      output_2: shape=(0,), dtype=float32
        []
      output_3: shape=(0,), dtype=float32
        []

Differences with previous version (13)
--------------------------------------

**SchemaDiff**: ``Split`` (domain ``'ai.onnx'``)

* old version: 13
* new version: 18
* breaking: no

**Attributes:**

* added 'num_outputs': type=INT; required=False; default=UNDEFINED

**Documentation:**

* line similarity: 0.00 (+5/-3 lines)

.. code-block:: diff

    --- Split v13
    +++ Split v18
    @@ -1,3 +1,5 @@
    -Split a tensor into a list of tensors, along the specified
    -'axis'. Lengths of the parts can be specified using input 'split'.
    -Otherwise, the tensor is split to equal sized parts.
    +Split a tensor into a list of tensors, along the specified 'axis'.
    +Either input 'split' or the attribute 'num_outputs' should be specified, but not both.
    +If the attribute 'num_outputs' is specified, then the tensor is split into equal sized parts.
    +If the tensor is not evenly splittable into `num_outputs`, the last chunk will be smaller.
    +If the input 'split' is specified, it indicates the sizes of each output in the split.

Version History
---------------

- :doc:`Version 13 <Split-13>`
- :doc:`Version 11 <Split-11>`
- :doc:`Version 2 <Split-2>`
- :doc:`Version 1 <Split-1>`