tf.contrib.estimator.add_metrics

Creates a new tf.estimator.Estimator which has given metrics.

tf.contrib.estimator.add_metrics(
    estimator, metric_fn
)

Example:

def my_auc(labels, predictions):
  return {'auc': tf.metrics.auc(labels, predictions['logistic'])}

estimator = tf.estimator.DNNClassifier(...)
estimator = tf.contrib.estimator.add_metrics(estimator, my_auc)
estimator.train(...)
estimator.evaluate(...)

Example usage of custom metric which uses features:

def my_auc(features, labels, predictions):
  return {'auc': tf.metrics.auc(
    labels, predictions['logistic'], weights=features['weight'])}

estimator = tf.estimator.DNNClassifier(...)
estimator = tf.contrib.estimator.add_metrics(estimator, my_auc)
estimator.train(...)
estimator.evaluate(...)