View source on GitHub |
A layer that produces a dense Tensor
based on given feature_columns
.
tf.compat.v1.keras.layers.DenseFeatures(
feature_columns, trainable=True, name=None, partitioner=None, **kwargs
)
Generally a single example in training data is described with
FeatureColumns. At the first layer of the model, this column-oriented data
should be converted to a single Tensor
.
This layer can be called multiple times with different features.
This is the V1 version of this layer that uses variable_scope's or
partitioner to create variables which works well with PartitionedVariables.
Variable scopes are deprecated in V2, so the V2 version uses name_scopes
instead. But currently that lacks support for partitioned variables. Use
this if you need partitioned variables. Use the partitioner argument if you
have a Keras model and uses
tf.compat.v1.keras.estimator.model_to_estimator
for training.
Example:
price = tf.feature_column.numeric_column('price')
keywords_embedded = tf.feature_column.embedding_column(
tf.feature_column.categorical_column_with_hash_bucket("keywords", 10K),
dimension=16)
columns = [price, keywords_embedded, ...]
partitioner = tf.compat.v1.fixed_size_partitioner(num_shards=4)
feature_layer = tf.compat.v1.keras.layers.DenseFeatures(
feature_columns=columns, partitioner=partitioner)
features = tf.io.parse_example(
..., features=tf.feature_column.make_parse_example_spec(columns))
dense_tensor = feature_layer(features)
for units in [128, 64, 32]:
dense_tensor = tf.compat.v1.keras.layers.Dense(
units, activation='relu')(dense_tensor)
prediction = tf.compat.v1.keras.layers.Dense(1)(dense_tensor)
Raises | |
---|---|
ValueError
|
if an item in feature_columns is not a DenseColumn .
|