InstanceNormalization - 1 vs 6

Files changed (1) hide show
  1. InstanceNormalization1 → InstanceNormalization6 +7 -6
InstanceNormalization1 → InstanceNormalization6 RENAMED
@@ -1 +1 @@
1
1
  Carries out instance normalization as described in the paper
2
2
  https://arxiv.org/abs/1607.08022.
3
3
  y = scale * (x - mean) / sqrt(variance + epsilon) + B,
4
4
  where mean and variance are computed per instance per channel.
5
5
  **Attributes**
6
- * **consumed_inputs**:
7
- legacy optimization attribute.
8
6
  * **epsilon**:
9
- The epsilon value to use to avoid division by zero, default is
7
+ The epsilon value to use to avoid division by zero.
10
- 1e-5f.
11
8
  **Inputs**
12
9
  * **input** (heterogeneous) - **T**:
10
+ Input data tensor from the previous operator; dimensions for image
11
+ case are (N x C x H x W), where N is the batch size, C is the number
12
+ of channels, and H and W are the height and the width of the data.
13
- The input 4-dimensional tensor of shape NCHW.
13
+ For non image case, the dimensions are in the form of (N x C x D1 x
14
+ D2 ... Dn), where N is the batch size.
14
15
  * **scale** (heterogeneous) - **T**:
15
16
  The input 1-dimensional scale tensor of size C.
16
17
  * **B** (heterogeneous) - **T**:
17
18
  The input 1-dimensional bias tensor of size C.
18
19
  **Outputs**
19
20
  * **output** (heterogeneous) - **T**:
20
- The output 4-dimensional tensor of the same shape as input.
21
+ The output tensor of the same shape as input.
21
22
  **Type Constraints**
22
23
  * **T** in (
23
24
  tensor(double),
24
25
  tensor(float),
25
26
  tensor(float16)
26
27
  ):
27
28
  Constrain input and output types to float tensors.