Sub - 1 vs 6#

Next section compares an older to a newer version of the same operator after both definition are converted into markdown text. Green means an addition to the newer version, red means a deletion. Anything else is unchanged.

Files changed (1) hide show
  1. Sub1 → Sub6 +4 -6
Sub1 → Sub6 RENAMED
@@ -1 +1 @@
1
1
  Performs element-wise binary subtraction (with limited broadcast support).
2
2
  If necessary the right-hand-side argument will be broadcasted to match the
3
3
  shape of left-hand-side argument. When broadcasting is specified, the second
4
4
  tensor can either be of element size 1 (including a scalar tensor and any
5
5
  tensor with rank equal to or smaller than the first tensor), or having its
6
6
  shape as a contiguous subset of the first tensor's shape. The starting of the
7
7
  mutually equal shape is specified by the argument "axis", and if it is not set,
8
8
  suffix matching is assumed. 1-dim expansion doesn't work yet.
9
9
  For example, the following tensor shapes are supported (with broadcast=1):
10
10
  shape(A) = (2, 3, 4, 5), shape(B) = (,), i.e. B is a scalar tensor
11
11
  shape(A) = (2, 3, 4, 5), shape(B) = (1, 1), i.e. B is an 1-element tensor
12
12
  shape(A) = (2, 3, 4, 5), shape(B) = (5,)
13
13
  shape(A) = (2, 3, 4, 5), shape(B) = (4, 5)
14
14
  shape(A) = (2, 3, 4, 5), shape(B) = (3, 4), with axis=1
15
15
  shape(A) = (2, 3, 4, 5), shape(B) = (2), with axis=0
16
16
  Attribute broadcast=1 needs to be passed to enable broadcasting.
17
17
  **Attributes**
18
18
  * **axis**:
19
19
  If set, defines the broadcast dimensions. See doc for details.
20
20
  * **broadcast**:
21
21
  Pass 1 to enable broadcasting
22
+ * **consumed_inputs**:
23
+ legacy optimization attribute.
22
24
  **Inputs**
23
25
  * **A** (heterogeneous) - **T**:
24
26
  First operand, should share the type with the second operand.
25
27
  * **B** (heterogeneous) - **T**:
26
28
  Second operand. With broadcasting can be of smaller size than A. If
27
29
  broadcasting is disabled it should be of the same size.
28
30
  **Outputs**
29
31
  * **C** (heterogeneous) - **T**:
30
32
  Result, has same dimensions and type as A
31
33
  **Type Constraints**
32
34
  * **T** in (
33
35
  tensor(double),
34
36
  tensor(float),
35
- tensor(float16),
37
+ tensor(float16)
36
- tensor(int32),
37
- tensor(int64),
38
- tensor(uint32),
39
- tensor(uint64)
40
38
  ):
41
- Constrain input and output types to high-precision numeric tensors.? ^^^^^^^^^^^^ ^^^^^^^^^
39
+ Constrain input and output types to float tensors.? ^^ ^^