Norm¶
norm
¶
PixelNorm
¶
Bases: Module
Pixel-wise feature vector normalization (ProGAN / StyleGAN).
Normalizes each feature vector to unit length along the channel dimension.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
eps
|
float
|
Epsilon for numerical stability. |
1e-08
|
ScaleNorm
¶
Bases: Module
Scale normalization (L2 normalize then scale by a learned parameter).
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
dim
|
int
|
Feature dimension. |
required |
eps
|
float
|
Epsilon for numerical stability. |
1e-08
|
pixel_norm
¶
Miscellaneous normalization layers.
PixelNorm
¶
Bases: Module
Pixel-wise feature vector normalization (ProGAN / StyleGAN).
Normalizes each feature vector to unit length along the channel dimension.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
eps
|
float
|
Epsilon for numerical stability. |
1e-08
|
ScaleNorm
¶
Bases: Module
Scale normalization (L2 normalize then scale by a learned parameter).
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
dim
|
int
|
Feature dimension. |
required |
eps
|
float
|
Epsilon for numerical stability. |
1e-08
|