Coverage for /pythoncovmergedfiles/medio/medio/usr/local/lib/python3.8/site-packages/keras/src/feature_column/dense_features.py: 45%

42 statements  

« prev     ^ index     » next       coverage.py v7.4.0, created at 2024-01-03 07:57 +0000

1# Copyright 2019 The TensorFlow Authors. All Rights Reserved. 

2# 

3# Licensed under the Apache License, Version 2.0 (the "License"); 

4# you may not use this file except in compliance with the License. 

5# You may obtain a copy of the License at 

6# 

7# http://www.apache.org/licenses/LICENSE-2.0 

8# 

9# Unless required by applicable law or agreed to in writing, software 

10# distributed under the License is distributed on an "AS IS" BASIS, 

11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 

12# See the License for the specific language governing permissions and 

13# limitations under the License. 

14# ============================================================================== 

15"""A layer that produces a dense `Tensor` based on given `feature_columns`.""" 

16 

17from __future__ import absolute_import 

18from __future__ import division 

19from __future__ import print_function 

20 

21import json 

22 

23import tensorflow.compat.v2 as tf 

24 

25from keras.src import backend 

26from keras.src.feature_column import base_feature_layer as kfc 

27from keras.src.saving.legacy.saved_model import json_utils 

28 

29# isort: off 

30from tensorflow.python.util.tf_export import keras_export 

31 

32 

33@keras_export(v1=["keras.layers.DenseFeatures"]) 

34class DenseFeatures(kfc._BaseFeaturesLayer): 

35 """A layer that produces a dense `Tensor` based on given `feature_columns`. 

36 

37 Generally a single example in training data is described with 

38 FeatureColumns. At the first layer of the model, this column-oriented data 

39 should be converted to a single `Tensor`. 

40 

41 This layer can be called multiple times with different features. 

42 

43 This is the V1 version of this layer that uses variable_scope's or 

44 partitioner to create variables which works well with PartitionedVariables. 

45 Variable scopes are deprecated in V2, so the V2 version uses name_scopes 

46 instead. But currently that lacks support for partitioned variables. Use 

47 this if you need partitioned variables. Use the partitioner argument if you 

48 have a Keras model and uses 

49 `tf.compat.v1.keras.estimator.model_to_estimator` for training. 

50 

51 Example: 

52 

53 ```python 

54 price = tf.feature_column.numeric_column('price') 

55 keywords_embedded = tf.feature_column.embedding_column( 

56 tf.feature_column.categorical_column_with_hash_bucket("keywords", 10K), 

57 dimension=16) 

58 columns = [price, keywords_embedded, ...] 

59 partitioner = tf.compat.v1.fixed_size_partitioner(num_shards=4) 

60 feature_layer = tf.compat.v1.keras.layers.DenseFeatures( 

61 feature_columns=columns, partitioner=partitioner) 

62 

63 features = tf.io.parse_example( 

64 ..., features=tf.feature_column.make_parse_example_spec(columns)) 

65 dense_tensor = feature_layer(features) 

66 for units in [128, 64, 32]: 

67 dense_tensor = tf.compat.v1.keras.layers.Dense( 

68 units, activation='relu')(dense_tensor) 

69 prediction = tf.compat.v1.keras.layers.Dense(1)(dense_tensor) 

70 ``` 

71 """ 

72 

73 def __init__( 

74 self, 

75 feature_columns, 

76 trainable=True, 

77 name=None, 

78 partitioner=None, 

79 **kwargs 

80 ): 

81 """Constructs a DenseFeatures layer. 

82 

83 Args: 

84 feature_columns: An iterable containing the FeatureColumns to use as 

85 inputs to your model. All items should be instances of classes 

86 derived from `DenseColumn` such as `numeric_column`, 

87 `embedding_column`, `bucketized_column`, `indicator_column`. If you 

88 have categorical features, you can wrap them with an 

89 `embedding_column` or `indicator_column`. 

90 trainable: Boolean, whether the layer's variables will be updated via 

91 gradient descent during training. 

92 name: Name to give to the DenseFeatures. 

93 partitioner: Partitioner for input layer. Defaults to `None`. 

94 **kwargs: Keyword arguments to construct a layer. 

95 

96 Raises: 

97 ValueError: if an item in `feature_columns` is not a `DenseColumn`. 

98 """ 

99 super().__init__( 

100 feature_columns=feature_columns, 

101 trainable=trainable, 

102 name=name, 

103 partitioner=partitioner, 

104 expected_column_type=tf.__internal__.feature_column.DenseColumn, 

105 **kwargs 

106 ) 

107 

108 @property 

109 def _is_feature_layer(self): 

110 return True 

111 

112 @property 

113 def _tracking_metadata(self): 

114 """String stored in metadata field in the SavedModel proto. 

115 

116 Returns: 

117 A serialized JSON storing information necessary for recreating this 

118 layer. 

119 """ 

120 metadata = json.loads(super()._tracking_metadata) 

121 metadata["_is_feature_layer"] = True 

122 return json.dumps(metadata, default=json_utils.get_json_type) 

123 

124 def _target_shape(self, input_shape, total_elements): 

125 return (input_shape[0], total_elements) 

126 

127 def call(self, features, cols_to_output_tensors=None, training=None): 

128 """Returns a dense tensor corresponding to the `feature_columns`. 

129 

130 Example usage: 

131 

132 >>> t1 = tf.feature_column.embedding_column( 

133 ... tf.feature_column.categorical_column_with_hash_bucket("t1", 2), 

134 ... dimension=8) 

135 >>> t2 = tf.feature_column.numeric_column('t2') 

136 >>> feature_layer = tf.compat.v1.keras.layers.DenseFeatures([t1, t2]) 

137 >>> features = {"t1": tf.constant(["a", "b"]), 

138 ... "t2": tf.constant([1, 2])} 

139 >>> dense_tensor = feature_layer(features, training=True) 

140 

141 Args: 

142 features: A mapping from key to tensors. `FeatureColumn`s look up via 

143 these keys. For example `numeric_column('price')` will look at 

144 'price' key in this dict. Values can be a `SparseTensor` or a 

145 `Tensor` depends on corresponding `FeatureColumn`. 

146 cols_to_output_tensors: If not `None`, this will be filled with a dict 

147 mapping feature columns to output tensors created. 

148 training: Python boolean or None, indicating whether to the layer is 

149 being run in training mode. This argument is passed to the call 

150 method of any `FeatureColumn` that takes a `training` argument. For 

151 example, if a `FeatureColumn` performed dropout, the column could 

152 expose a `training` argument to control whether the dropout should 

153 be applied. If `None`, becomes `tf.keras.backend.learning_phase()`. 

154 Defaults to `None`. 

155 

156 

157 Returns: 

158 A `Tensor` which represents input layer of a model. Its shape 

159 is (batch_size, first_layer_dimension) and its dtype is `float32`. 

160 first_layer_dimension is determined based on given `feature_columns`. 

161 

162 Raises: 

163 ValueError: If features are not a dictionary. 

164 """ 

165 if training is None: 

166 training = backend.learning_phase() 

167 if not isinstance(features, dict): 

168 raise ValueError( 

169 "We expected a dictionary here. Instead we got: ", features 

170 ) 

171 transformation_cache = ( 

172 tf.__internal__.feature_column.FeatureTransformationCache(features) 

173 ) 

174 output_tensors = [] 

175 for column in self._feature_columns: 

176 with backend.name_scope(column.name): 

177 try: 

178 tensor = column.get_dense_tensor( 

179 transformation_cache, 

180 self._state_manager, 

181 training=training, 

182 ) 

183 except TypeError: 

184 tensor = column.get_dense_tensor( 

185 transformation_cache, self._state_manager 

186 ) 

187 processed_tensors = self._process_dense_tensor(column, tensor) 

188 if cols_to_output_tensors is not None: 

189 cols_to_output_tensors[column] = processed_tensors 

190 output_tensors.append(processed_tensors) 

191 return self._verify_and_concat_tensors(output_tensors) 

192