BatchNormalization - 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Next section compares an older to a newer version of the same operator after both definition are converted into markdown text.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.
BatchNormalization1 → BatchNormalization6
RENAMED
|
@@ -1 +1 @@
|
|
|
1
1
|
Carries out batch normalization as described in the paper
|
|
2
2
|
https://arxiv.org/abs/1502.03167. Depending on the mode it is being run,
|
|
3
3
|
there are multiple cases for the number of outputs, which we list below:
|
|
4
4
|
Output case #1: Y, mean, var, saved_mean, saved_var (training mode)
|
|
5
5
|
Output case #2: Y (test mode)
|
|
6
6
|
**Attributes**
|
|
7
|
+
* **consumed_inputs** (required):
|
|
8
|
+
legacy optimization attribute.
|
|
7
9
|
* **epsilon**:
|
|
8
10
|
The epsilon value to use to avoid division by zero, default is
|
|
9
11
|
1e-5f.
|
|
10
12
|
* **is_test**:
|
|
11
13
|
If set to nonzero, run spatial batch normalization in test mode,
|
|
12
14
|
default is 0.
|
|
13
15
|
* **momentum**:
|
|
14
16
|
Factor used in computing the running mean and variance.e.g.,
|
|
15
17
|
running_mean = running_mean * momentum + mean * (1 - momentum),
|
|
16
18
|
default is 0.9f.
|
|
17
19
|
* **spatial**:
|
|
18
20
|
If true, compute the mean and variance across all spatial elements
|
|
19
21
|
If false, compute the mean and variance across per feature.Default
|
|
20
22
|
is 1.
|
|
21
23
|
**Inputs**
|
|
22
24
|
* **X** (heterogeneous) - **T**:
|
|
25
|
+
The input 4-dimensional tensor of shape NCHW.
|
|
23
|
-
Input data tensor from the previous operator; dimensions for image
|
|
24
|
-
case are (N x C x H x W), where N is the batch size, C is the number
|
|
25
|
-
of channels, and H and W are the height and the width of the data.
|
|
26
|
-
For non image case, the dimensions are in the form of (N x C x D1 x
|
|
27
|
-
D2 ... Dn), where N is the batch size.
|
|
28
26
|
* **scale** (heterogeneous) - **T**:
|
|
29
27
|
The scale as a 1-dimensional tensor of size C to be applied to the
|
|
30
28
|
output.
|
|
31
29
|
* **B** (heterogeneous) - **T**:
|
|
32
30
|
The bias as a 1-dimensional tensor of size C to be applied to the
|
|
33
31
|
output.
|
|
34
32
|
* **mean** (heterogeneous) - **T**:
|
|
35
33
|
The running mean (training) or the estimated mean (testing) as a
|
|
36
34
|
1-dimensional tensor of size C.
|
|
37
35
|
* **var** (heterogeneous) - **T**:
|
|
38
36
|
The running variance (training) or the estimated variance (testing)
|
|
39
37
|
as a 1-dimensional tensor of size C.
|
|
40
38
|
**Outputs**
|
|
41
39
|
Between 1 and 5 outputs.
|
|
42
40
|
* **Y** (heterogeneous) - **T**:
|
|
43
|
-
The output tensor of the same shape as X.
|
|
41
|
+
The output 4-dimensional tensor of the same shape as X.
|
|
44
42
|
* **mean** (optional, heterogeneous) - **T**:
|
|
45
43
|
The running mean after the BatchNormalization operator. Must be in-
|
|
46
44
|
place with the input mean. Should not be used for testing.
|
|
47
45
|
* **var** (optional, heterogeneous) - **T**:
|
|
48
46
|
The running variance after the BatchNormalization operator. Must be
|
|
49
47
|
in-place with the input var. Should not be used for testing.
|
|
50
48
|
* **saved_mean** (optional, heterogeneous) - **T**:
|
|
51
49
|
Saved mean used during training to speed up gradient computation.
|
|
52
50
|
Should not be used for testing.
|
|
53
51
|
* **saved_var** (optional, heterogeneous) - **T**:
|
|
54
52
|
Saved variance used during training to speed up gradient
|
|
55
53
|
computation. Should not be used for testing.
|
|
56
54
|
**Type Constraints**
|
|
57
55
|
* **T** in (
|
|
58
56
|
tensor(double),
|
|
59
57
|
tensor(float),
|
|
60
58
|
tensor(float16)
|
|
61
59
|
):
|
|
62
60
|
Constrain input and output types to float tensors.
|