def dump_model_parameters(model):
# now show the values of the initial values of all parameters
assert model
import keras.backend as K
print K.batch_get_value(model.weights)
def weights_match(model1, model2):
"""Determine whether the weights of two Keras models match.
Args:
model1: First model
model2: Second model
Returns:
result: True if there is a match
"""
# Get weights
weights_m1 = backend.batch_get_value(model1.weights)
weights_m2 = backend.batch_get_value(model2.weights)
# Evaluate equivalence
if all([np.all(w == ow) for w, ow in zip(weights_m1, weights_m2)]):
result = True
else:
result = False
return result
def _get_weight_names_and_values_from_symbolic(self, symbolic_weights):
import keras.backend as K # pylint: disable=import-error
weight_values = K.batch_get_value(symbolic_weights)
weight_names = [
(
str(w.name) if hasattr(w, 'name') and w.name \
else 'param_{}'.format(i)
)
for i, (w, val) in enumerate(
zip(symbolic_weights, weight_values)
)
]
return weight_names, weight_values
def XXXget_layer_weights(self):
"""
get the layers of the current model...
"""
result = {}
jlayer=0
for layer in self.model.layers:
symbolic_weights = layer.trainable_weights + layer.non_trainable_weights
weight_values = K.batch_get_value(symbolic_weights)
layer_list=[]
for i, (w, val) in enumerate(zip(symbolic_weights, weight_values)):
layer_list.append(val)
print("ADDING W",jlayer)
result[jlayer] = layer_list
jlayer += 1
return result
def apply_ema_weights(self):
"""备份原模型权重，然后将平均权重应用到模型上去。
"""
self.old_weights = K.batch_get_value(self.model.weights)
ema_weights = K.batch_get_value(self.ema_weights)
K.batch_set_value(zip(self.model.weights, ema_weights))
def initialize(self):
"""ema_weights初始化跟原模型初始化一致。
"""
self.old_weights = K.batch_get_value(self.model.weights)
K.batch_set_value(zip(self.ema_weights, self.old_weights))
def get_weights(self):
weights = []
for cell in self.cells:
if isinstance(cell, Layer):
weights += cell.weights
return K.batch_get_value(weights)
def get_weights(self):
# ref: https://github.com/keras-team/keras/blob/c10d24959b0ad615a21e671b180a1b2466d77a2b/keras/engine/base_layer.py#L21
params = self.weights
weights = K.batch_get_value(params)
return weights[0] * K.get_value(self.tree_weight), weights[1]
def reduce_model(model):
"""Returns a dict representing the state of the model.
The dict contains:
- the model's configuration (topology)
- the model's weights
- the model's optimizer's state (if any)
Thus the model can be reinstantiated in
the returned dict, without any of the code
used for model definition or training.
# Arguments
model: Keras model instance.
"""
def get_config(obj):
# if obj is a serializable Keras class instance
# e.g. optimizer, layer
if hasattr(obj, 'get_config'):
return {
'class_name': obj.__class__.__name__,
'config': obj.get_config()
}
return obj
attrs = {}
attrs['keras_version'] = str(keras_version).encode('utf8')
attrs['backend'] = K.backend().encode('utf8')
round_tripped_model = model_from_config(get_config(model))
attrs['model_config'] = get_config(round_tripped_model)
model_weights = {}
attrs['model_weights'] = model_weights
if legacy_models.needs_legacy_support(model):
model_layers = legacy_models.legacy_sequential_layers(model)
else:
model_layers = model.layers
model_weights['layer_names'] = [
layer.name.encode('utf8') for layer in model_layers
]
model_weights['backend'] = K.backend().encode('utf8')
model_weights['keras_version'] = str(keras_version).encode('utf8')
for layer in model_layers:
g = {}
model_weights[layer.name] = g
symbolic_weights = layer.weights
weight_values = K.batch_get_value(symbolic_weights)
weight_names = []
g['weight_names'] = weight_names
for i, (w, val) in enumerate(zip(symbolic_weights, weight_values)):
if hasattr(w, 'name') and w.name:
name = str(w.name)
else:
name = 'param_' + str(i)
weight_names.append(name)
g[name] = val
if hasattr(model, 'optimizer'):
if isinstance(model.optimizer, optimizers.TFOptimizer):
warnings.warn(
'TensorFlow optimizers do not '
'make it possible to access '
'optimizer attributes or optimizer state '
'after instantiation. '
'As a result, we cannot save the optimizer '
'as part of the model save file.'
'You will have to compile your model again after loading it. '
'Prefer using a Keras optimizer instead '
'(see keras.io/optimizers).')
else:
attrs['training_config'] = t.valmap(
get_config, {
'optimizer_config': model.optimizer,
'loss': model.loss,
'metrics': model.metrics,
'sample_weight_mode': model.sample_weight_mode,
'loss_weights': model.loss_weights,
})
# Save optimizer weights.
symbolic_weights = getattr(model.optimizer, 'weights')
if symbolic_weights:
optimizer_weights = {}
attrs['optimizer_weigts'] = optimizer_weights
weight_values = K.batch_get_value(symbolic_weights)
for i, (w,
val) in enumerate(zip(symbolic_weights,
weight_values)):
# Default values of symbolic_weights is /variable for theano
if K.backend() == 'theano':
if hasattr(w, 'name') and w.name != "/variable":
name = str(w.name)
else:
name = 'param_' + str(i)
else:
if hasattr(w, 'name') and w.name:
name = str(w.name)
else:
name = 'param_' + str(i)
optimizer_weights[name] = val
return attrs