LpPool - 2 vs 11#
Next section compares an older to a newer version of the same operator after both definition are converted into markdown text.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.
- LpPool2 → LpPool11 +5 -8
LpPool2 → LpPool11
RENAMED
|
@@ -1 +1 @@
|
|
|
1
1
|
LpPool consumes an input tensor X and applies Lp pooling across
|
|
2
2
|
the tensor according to kernel sizes, stride sizes, and pad lengths.
|
|
3
3
|
Lp pooling consisting of computing the Lp norm on all values of a subset
|
|
4
4
|
of the input tensor according to the kernel size and downsampling the
|
|
5
5
|
data into the output tensor Y for further processing.
|
|
6
6
|
**Attributes**
|
|
7
7
|
* **auto_pad**:
|
|
8
8
|
auto_pad must be either NOTSET, SAME_UPPER, SAME_LOWER or VALID.
|
|
9
9
|
Where default value is NOTSET, which means explicit padding is used.
|
|
10
|
-
SAME_UPPER or SAME_LOWER mean pad the input so that
|
|
10
|
+
SAME_UPPER or SAME_LOWER mean pad the input so that the output
|
|
11
|
+
spatial size match the input.In case of odd number add the extra
|
|
12
|
+
padding at the end for SAME_UPPER and at the beginning for
|
|
13
|
+
SAME_LOWER. VALID mean no padding.
|
|
11
|
-
= ceil(input_shape[i] / strides[i]) for each axis i. The padding
|
|
12
|
-
is split between the two sides equally or almost equally (depending
|
|
13
|
-
on whether it is even or odd). In case the padding is an odd number,
|
|
14
|
-
the extra padding is added at the end for SAME_UPPER and at the
|
|
15
|
-
beginning for SAME_LOWER.
|
|
16
14
|
* **kernel_shape** (required):
|
|
17
15
|
The size of the kernel along each axis.
|
|
18
16
|
* **p**:
|
|
19
17
|
p value of the Lp norm used to pool over the input data.
|
|
20
18
|
* **pads**:
|
|
21
19
|
Padding for the beginning and ending along each spatial axis, it can
|
|
22
20
|
take any value greater than or equal to 0. The value represent the
|
|
23
21
|
number of pixels added to the beginning and end part of the
|
|
24
22
|
corresponding axis. pads format should be as follow [x1_begin,
|
|
25
23
|
x2_begin...x1_end, x2_end,...], where xi_begin the number of pixels
|
|
26
24
|
added at the beginning of axis i and xi_end, the number of pixels
|
|
27
25
|
added at the end of axis i. This attribute cannot be used
|
|
28
26
|
simultaneously with auto_pad attribute. If not present, the padding
|
|
29
27
|
defaults to 0 along start and end of each spatial axis.
|
|
30
28
|
* **strides**:
|
|
31
|
-
Stride along each spatial axis. If not present, the stride defaults
|
|
32
|
-
|
|
29
|
+
Stride along each spatial axis.
|
|
33
30
|
**Inputs**
|
|
34
31
|
* **X** (heterogeneous) - **T**:
|
|
35
32
|
Input data tensor from the previous operator; dimensions for image
|
|
36
33
|
case are (N x C x H x W), where N is the batch size, C is the number
|
|
37
34
|
of channels, and H and W are the height and the width of the data.
|
|
38
35
|
For non image case, the dimensions are in the form of (N x C x D1 x
|
|
39
36
|
D2 ... Dn), where N is the batch size.
|
|
40
37
|
**Outputs**
|
|
41
38
|
* **Y** (heterogeneous) - **T**:
|
|
42
39
|
Output data tensor from Lp pooling across the input tensor.
|
|
43
40
|
Dimensions will vary based on various kernel, stride, and pad sizes.
|
|
44
41
|
**Type Constraints**
|
|
45
42
|
* **T** in (
|
|
46
43
|
tensor(double),
|
|
47
44
|
tensor(float),
|
|
48
45
|
tensor(float16)
|
|
49
46
|
):
|
|
50
47
|
Constrain input and output types to float tensors.
|