def get_config(self):
from tensorflow.python.feature_column.feature_column_lib import serialize_feature_columns
config = {
"columns": serialize_feature_columns(self.columns),
"cross_columns": serialize_feature_columns(self.cross_columns)
}
base_config = super(Poly2Model, self).get_config()
return {**base_config, **config}
def get_config(self):
from tensorflow.python.feature_column.feature_column_lib import serialize_feature_columns
config = {
"output_dim": self.output_dim,
"user_columns": serialize_feature_columns(self.user_columns),
"item_columns": serialize_feature_columns(self.item_columns),
"hidden_units": tf.keras.utils.serialize_keras_object(self.hidden_units),
"dropout": self.dropout
}
base_config = super(NeuralCfModel, self).get_config()
return {**base_config, **config}
def get_config(self):
from tensorflow.python.feature_column.feature_column_lib import serialize_feature_columns
config = {
"units":
self.units,
"linear_columns":
serialize_feature_columns(self.linear_columns),
"dnn_columns":
serialize_feature_columns(self.dnn_columns),
"hidden_units":
tf.keras.utils.serialize_keras_object(self.hidden_units),
"dropout":
self.dropout
}
base_config = super(WideDeepModel, self).get_config()
return {**base_config, **config}
def get_config(self):
"""Returns a dictionary with the config of the model."""
config = {'name': self.name}
config['rnn_layer'] = {
'class_name': self._rnn_layer.__class__.__name__,
'config': self._rnn_layer.get_config()
}
config['units'] = self._logits_layer.units
config['return_sequences'] = self._return_sequences
config['activation'] = activations.serialize(self._logits_layer.activation)
config['sequence_feature_columns'] = fc.serialize_feature_columns(
self._sequence_feature_columns)
config['context_feature_columns'] = (
fc.serialize_feature_columns(self._context_feature_columns)
if self._context_feature_columns else None)
return config
def get_config(self):
from tensorflow.python.feature_column.feature_column_lib import serialize_feature_columns
config = {
"latent_dim": self.latent_dim,
"columns": serialize_feature_columns(self.columns)
}
base_config = super(FMModel, self).get_config()
return {**base_config, **config}
def get_config(self):
from tensorflow.python.feature_column.feature_column_lib import serialize_feature_columns
config = {
"columns": serialize_feature_columns(self.columns),
"l2_factor": self.l2_factor
}
base_config = super(LinearModel, self).get_config()
return {**base_config, **config}
def get_config(self):
from tensorflow.python.feature_column.feature_column_lib import serialize_feature_columns
config = {
"output_dim":
self.output_dim,
"feature_columns":
serialize_feature_columns(self.feature_columns),
"residual_units":
tf.keras.utils.serialize_keras_object(self.residual_units)
}
base_config = super(DeepCrossingModel, self).get_config()
return {**base_config, **config}
def get_config(self):
from tensorflow.python.feature_column.feature_column_lib import serialize_feature_columns
config = {
"output_dim": self.output_dim,
"feature_columns": serialize_feature_columns(self.feature_columns),
"hidden_units": tf.keras.utils.serialize_keras_object(self.hidden_units),
"activation_fn": self.activation_fn,
"dropout": self.dropout,
"batch_norm": self.batch_norm
}
base_config = super(DNNModel, self).get_config()
return {**base_config, **config}
def get_config(self):
from tensorflow.python.feature_column.feature_column_lib import serialize_feature_column, \
serialize_feature_columns
config = {
"average_score": self.average_score,
"latent_dim": self.latent_dim,
"user_column": serialize_feature_column(self.user_column),
"item_column": serialize_feature_column(self.item_column),
"user_history_columns": serialize_feature_columns(self.user_history_columns),
"l2_factor_bias": self.l2_factor_bias,
"l2_factor_embedding": self.l2_factor_embedding
}
base_config = super(SVDPlusPlusModel, self).get_config()
return {**base_config, **config}
def get_config(self):
from tensorflow.python.feature_column.feature_column_lib import serialize_feature_columns
config = {
"output_dim":
self.output_dim,
"embedding_size":
self.embedding_size,
"feature_columns":
serialize_feature_columns(self.feature_columns),
"product_type":
self.product_type,
"hidden_units":
tf.keras.utils.serialize_keras_object(self.hidden_units),
"dropout":
self.dropout
}
base_config = super(PNNModel, self).get_config()
return {**base_config, **config}
def serialize_feature_columns(feature_columns):
"""Serializes feature columns to a dict of class name and config.
This serialization is required to support for SavedModel using model.save()
in Keras. The serialization is similar to that of
`tf.keras.layers.DenseFeatures`, which also has feature columns in it's
constructor.
Args:
feature_columns: (dict) feature names to feature columns.
Returns:
A dict mapping feature names to serialized feature columns.
"""
if not feature_columns:
return {}
feature_column_configs = {}
sorted_name_to_feature_columns = sorted(six.iteritems(feature_columns))
sorted_names, sorted_feature_columns = zip(*sorted_name_to_feature_columns)
sorted_configs = fc.serialize_feature_columns(sorted_feature_columns)
feature_column_configs = dict(zip(sorted_names, sorted_configs))
return feature_column_configs