tf.contrib.estimator.dnn_logit_fn_builder

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

Defined in tensorflow/python/estimator/canned/dnn.py.

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.

Returns:

A logit_fn (see below).

Raises:

  • ValueError: If units is not an int.