github/DeepFaceLab
1
0
Fork 0
mirror of https://github.com/iperov/DeepFaceLab.git synced 2025-07-06 21:12:07 -07:00
Code Issues Projects Releases Packages Wiki Activity Actions

revert back Adam

Browse source
This commit is contained in:
iperov 2019-03-11 21:52:36 +04:00
parent e4637336ef
commit ee8dbcbc35
3 changed files with 56 additions and 66 deletions
Download patch file Download diff file Expand all files Collapse all files

62
nnlib/nnlib.py
View file

@ -71,8 +71,8 @@ ZeroPadding2D = keras.layers.ZeroPadding2D
RandomNormal = keras.initializers.RandomNormal
Model = keras.models.Model
#Adam = keras.optimizers.Adam
Adam = nnlib.Adam
Adam = keras.optimizers.Adam
FastAdam = nnlib.FastAdam
modelify = nnlib.modelify
gaussian_blur = nnlib.gaussian_blur
@ -434,21 +434,16 @@ NLayerDiscriminator = nnlib.NLayerDiscriminator
return dict(list(base_config.items()) + list(config.items()))
nnlib.Scale = Scale
class Adam(keras.optimizers.Optimizer):
def __init__(self, lr=0.001, beta_1=0.9, beta_2=0.999,
epsilon=None, decay=0., amsgrad=False, iterations=0, **kwargs):
super(Adam, self).__init__(**kwargs)
class FastAdam(keras.optimizers.Optimizer):
def __init__(self, lr=0.001, beta_1=0.9, beta_2=0.999, iterations=0, **kwargs):
super(FastAdam, self).__init__(**kwargs)
with K.name_scope(self.__class__.__name__):
self.iterations = K.variable(iterations, dtype='int64', name='iterations')
self.lr = K.variable(lr, name='lr')
self.beta_1 = K.variable(beta_1, name='beta_1')
self.beta_2 = K.variable(beta_2, name='beta_2')
self.decay = K.variable(decay, name='decay')
if epsilon is None:
epsilon = K.epsilon()
self.epsilon = epsilon
self.initial_decay = decay
self.amsgrad = amsgrad
self.epsilon = K.epsilon()
@keras.legacy.interfaces.legacy_get_updates_support
def get_updates(self, loss, params):
@ -456,34 +451,16 @@ NLayerDiscriminator = nnlib.NLayerDiscriminator
self.updates = [K.update_add(self.iterations, 1)]
lr = self.lr
if self.initial_decay > 0:
lr = lr * (1. / (1. + self.decay * K.cast(self.iterations,
K.dtype(self.decay))))
t = K.cast(self.iterations, K.floatx()) + 1
lr_t = lr * (K.sqrt(1. - K.pow(self.beta_2, t)) /
(1. - K.pow(self.beta_1, t)))
(1. - K.pow(self.beta_1, t)))
self.weights = [self.iterations]
ms = [K.zeros(K.int_shape(p), dtype=K.dtype(p)) for p in params]
vs = [K.zeros(K.int_shape(p), dtype=K.dtype(p)) for p in params]
if self.amsgrad:
vhats = [K.zeros(K.int_shape(p), dtype=K.dtype(p)) for p in params]
else:
vhats = [K.zeros(1) for _ in params]
self.weights = [self.iterations] + ms + vs + vhats
for p, g in zip(params, grads):
for p, g, m, v, vhat in zip(params, grads, ms, vs, vhats):
m_t = (self.beta_1 * m) + (1. - self.beta_1) * g
v_t = (self.beta_2 * v) + (1. - self.beta_2) * K.square(g)
if self.amsgrad:
vhat_t = K.maximum(vhat, v_t)
p_t = p - lr_t * m_t / (K.sqrt(vhat_t) + self.epsilon)
self.updates.append(K.update(vhat, vhat_t))
else:
p_t = p - lr_t * m_t / (K.sqrt(v_t) + self.epsilon)
self.updates.append(K.update(m, m_t))
self.updates.append(K.update(v, v_t))
m_t = (1. - self.beta_1) * g
v_t = (1. - self.beta_2) * K.square(g)
p_t = p - lr_t * m_t / (K.sqrt(v_t) + self.epsilon)
new_p = p_t
# Apply constraints.
@ -497,15 +474,14 @@ NLayerDiscriminator = nnlib.NLayerDiscriminator
config = {'iterations': int(K.get_value(self.iterations)),
'lr': float(K.get_value(self.lr)),
'beta_1': float(K.get_value(self.beta_1)),
'beta_2': float(K.get_value(self.beta_2)),
'decay': float(K.get_value(self.decay)),
'epsilon': self.epsilon,
'amsgrad': self.amsgrad}
base_config = super(Adam, self).get_config()
'beta_2': float(K.get_value(self.beta_2))
}
base_config = super(FastAdam, self).get_config()
return dict(list(base_config.items()) + list(config.items()))
nnlib.Adam = Adam
'''
nnlib.FastAdam = FastAdam
'''
not implemented in plaidML
class ReflectionPadding2D(keras.layers.Layer):
def __init__(self, padding=(1, 1), **kwargs):