CumSum#

  • Domain: ai.onnx

  • Since version: 14

Performs cumulative sum of the input elements along the given axis. By default, it will do the sum inclusively meaning the first element is copied as is. Through an exclusive attribute, this behavior can change to exclude the first element. It can also perform summation in the opposite direction of the axis. For that, set reverse attribute to 1.

Example:

input_x = [1, 2, 3]
axis=0
output = [1, 3, 6]
exclusive=1
output = [0, 1, 3]
exclusive=0
reverse=1
output = [6, 5, 3]
exclusive=1
reverse=1
output = [5, 3, 0]

Inputs

  • x (T): An input tensor that is to be processed.

  • axis (T2): A 0-D tensor. Must be in the range [-rank(x), rank(x)-1]. Negative value means counting dimensions from the back.

Outputs

  • y (T): Output tensor of the same type as ‘x’ with cumulative sums of the x’s elements

Attributes

  • exclusive (int): If set to 1 will return exclusive sum in which the top element is not included. In other terms, if set to 1, the j-th output element would be the sum of the first (j-1) elements. Otherwise, it would be the sum of the first j elements.

  • reverse (int): If set to 1 will perform the sums in reverse direction.

Type Constraints

  • T: Constrain input and output types to numeric tensors. Allowed types: tensor(bfloat16), tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint32), tensor(uint64).

  • T2: axis tensor can be int32 or int64 only Allowed types: tensor(int32), tensor(int64).

Examples#

test_cumsum_1d

Node:
  CumSum(x, axis) -> (y)

Inputs:
  x: shape=(5,), dtype=float64
    [1., 2., 3., 4., 5.]
  axis: shape=(), dtype=int32
    0

Outputs:
  y: shape=(5,), dtype=float64
    [ 1.,  3.,  6., 10., 15.]

test_cumsum_1d_exclusive

Node:
  CumSum(x, axis) -> (y)
  Attributes:
    exclusive = 1

Inputs:
  x: shape=(5,), dtype=float64
    [1., 2., 3., 4., 5.]
  axis: shape=(), dtype=int32
    0

Outputs:
  y: shape=(5,), dtype=float64
    [ 0.,  1.,  3.,  6., 10.]

test_cumsum_1d_int32_exclusive

Node:
  CumSum(x, axis) -> (y)
  Attributes:
    exclusive = 1

Inputs:
  x: shape=(5,), dtype=int32
    [1, 2, 3, 4, 5]
  axis: shape=(), dtype=int32
    0

Outputs:
  y: shape=(5,), dtype=int32
    [ 0,  1,  3,  6, 10]

test_cumsum_1d_reverse

Node:
  CumSum(x, axis) -> (y)
  Attributes:
    reverse = 1

Inputs:
  x: shape=(5,), dtype=float64
    [1., 2., 3., 4., 5.]
  axis: shape=(), dtype=int32
    0

Outputs:
  y: shape=(5,), dtype=float64
    [15., 14., 12.,  9.,  5.]

test_cumsum_1d_reverse_exclusive

Node:
  CumSum(x, axis) -> (y)
  Attributes:
    exclusive = 1
    reverse = 1

Inputs:
  x: shape=(5,), dtype=float64
    [1., 2., 3., 4., 5.]
  axis: shape=(), dtype=int32
    0

Outputs:
  y: shape=(5,), dtype=float64
    [14., 12.,  9.,  5.,  0.]

test_cumsum_2d_axis_0

Node:
  CumSum(x, axis) -> (y)

Inputs:
  x: shape=(2, 3), dtype=float64
    [[1., 2., 3.],
     [4., 5., 6.]]
  axis: shape=(), dtype=int32
    0

Outputs:
  y: shape=(2, 3), dtype=float64
    [[1., 2., 3.],
     [5., 7., 9.]]

test_cumsum_2d_axis_1

Node:
  CumSum(x, axis) -> (y)

Inputs:
  x: shape=(2, 3), dtype=float64
    [[1., 2., 3.],
     [4., 5., 6.]]
  axis: shape=(), dtype=int32
    1

Outputs:
  y: shape=(2, 3), dtype=float64
    [[ 1.,  3.,  6.],
     [ 4.,  9., 15.]]

test_cumsum_2d_int32

Node:
  CumSum(x, axis) -> (y)

Inputs:
  x: shape=(2, 3), dtype=int32
    [[1, 2, 3],
     [4, 5, 6]]
  axis: shape=(), dtype=int32
    0

Outputs:
  y: shape=(2, 3), dtype=int32
    [[1, 2, 3],
     [5, 7, 9]]

test_cumsum_2d_negative_axis

Node:
  CumSum(x, axis) -> (y)

Inputs:
  x: shape=(2, 3), dtype=float64
    [[1., 2., 3.],
     [4., 5., 6.]]
  axis: shape=(), dtype=int64
    -1

Outputs:
  y: shape=(2, 3), dtype=float64
    [[ 1.,  3.,  6.],
     [ 4.,  9., 15.]]

Differences with previous version (11)#

SchemaDiff: CumSum (domain 'ai.onnx')

  • old version: 11

  • new version: 14

  • breaking: no

Type constraints:

  • changed ‘T’: added types: [‘tensor(bfloat16)’, ‘tensor(float16)’]

Version History#