github/DeepFaceLab
1
0
Fork 0
mirror of https://github.com/iperov/DeepFaceLab.git synced 2025-07-10 15:23:32 -07:00
Code Issues Projects Releases Packages Wiki Activity Actions

adding kernel initializer option to FUNITAdain block

Browse source
This commit is contained in:
Colombo 2019-09-24 21:30:28 +04:00
parent deeb98474b
commit 4a2203cc35
2 changed files with 6 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

4
nnlib/FUNIT.py
View file

@ -258,10 +258,10 @@ class FUNIT(object):
inp, mlp = input inp, mlp = input
x = inp x = inp
x = Conv2D(dim, 3, strides=1, padding='valid')(ZeroPadding2D(1)(x)) x = Conv2D(dim, 3, strides=1, padding='valid')(ZeroPadding2D(1)(x))
x = FUNITAdain()([x,mlp]) x = FUNITAdain(kernel_initializer='he_normal')([x,mlp])
x = ReLU()(x) x = ReLU()(x)
x = Conv2D(dim, 3, strides=1, padding='valid')(ZeroPadding2D(1)(x)) x = Conv2D(dim, 3, strides=1, padding='valid')(ZeroPadding2D(1)(x))
x = FUNITAdain()([x,mlp]) x = FUNITAdain(kernel_initializer='he_normal')([x,mlp])
return Add()([x,inp]) return Add()([x,inp])
return func return func

7
nnlib/nnlib.py
View file

@ -522,10 +522,11 @@ NLayerDiscriminator = nnlib.NLayerDiscriminator
so we don't need to slice outter MLP block and assign weights every call, just pass MLP inside. so we don't need to slice outter MLP block and assign weights every call, just pass MLP inside.
also size of dense blocks is calculated automatically also size of dense blocks is calculated automatically
""" """
def __init__(self, axis=-1, epsilon=1e-5, momentum=0.99, **kwargs): def __init__(self, axis=-1, epsilon=1e-5, momentum=0.99, kernel_initializer='glorot_uniform', **kwargs):
self.axis = axis self.axis = axis
self.epsilon = epsilon self.epsilon = epsilon
self.momentum = momentum self.momentum = momentum
self.kernel_initializer = kernel_initializer
super(FUNITAdain, self).__init__(**kwargs) super(FUNITAdain, self).__init__(**kwargs)
def build(self, input_shape): def build(self, input_shape):
@ -533,9 +534,9 @@ NLayerDiscriminator = nnlib.NLayerDiscriminator
x, mlp = input_shape x, mlp = input_shape
units = x[self.axis] units = x[self.axis]
self.kernel1 = self.add_weight(shape=(units, units), initializer='he_normal', name='kernel1') self.kernel1 = self.add_weight(shape=(units, units), initializer=self.kernel_initializer, name='kernel1')
self.bias1 = self.add_weight(shape=(units,), initializer='zeros', name='bias1') self.bias1 = self.add_weight(shape=(units,), initializer='zeros', name='bias1')
self.kernel2 = self.add_weight(shape=(units, units), initializer='he_normal', name='kernel2') self.kernel2 = self.add_weight(shape=(units, units), initializer=self.kernel_initializer, name='kernel2')
self.bias2 = self.add_weight(shape=(units,), initializer='zeros', name='bias2') self.bias2 = self.add_weight(shape=(units,), initializer='zeros', name='bias2')
self.built = True self.built = True