tf.estimator.experimental.dnn_logit_fn_builder

Aliases:

• tf.contrib.estimator.dnn_logit_fn_builder
• tf.estimator.experimental.dnn_logit_fn_builder

tf.estimator.experimental.dnn_logit_fn_builder(
    units,
    hidden_units,
    feature_columns,
    activation_fn,
    dropout,
    input_layer_partitioner,
    batch_norm
)

Function builder for a dnn logit_fn.

Args:

• units: An int indicating the dimension of the logit layer. In the MultiHead case, this should be the sum of all component Heads' logit dimensions.
• hidden_units: Iterable of integer number of hidden units per layer.
• feature_columns: Iterable of feature_column._FeatureColumn model inputs.
• activation_fn: Activation function applied to each layer.
• dropout: When not None, the probability we will drop out a given coordinate.
• input_layer_partitioner: Partitioner for input layer.
• batch_norm: Whether to use batch normalization after each hidden layer.

Returns:

A logit_fn (see below).

Raises:

• ValueError: If units is not an int.