def evaluate()

in adanet/core/evaluator.py [0:0]

• 25 lines of code
• 13 McCabe index (conditional complexity)

  def evaluate(self, sess, ensemble_metrics):
    """Evaluates the given AdaNet objectives on the data from `input_fn`.

    The candidates are fed the same batches of features and labels as
    provided by `input_fn`, and their losses are computed and summed over
    `steps` batches.

    Args:
      sess: `Session` instance with most recent variable values loaded.
      ensemble_metrics: A list dictionaries of `tf.metrics` for each candidate
        ensemble.

    Returns:
      List of evaluated metrics.
    """

    evals_completed = 0
    if self.steps is None:
      logging_frequency = 1000
    elif self.steps < 10:
      logging_frequency = 1
    else:
      logging_frequency = math.floor(self.steps / 10.)

    objective_metrics = [em[self._metric_name] for em in ensemble_metrics]

    sess.run(tf_compat.v1.local_variables_initializer())
    while True:
      if self.steps is not None and evals_completed == self.steps:
        break
      try:
        evals_completed += 1
        if (evals_completed % logging_frequency == 0 or
            self.steps == evals_completed):
          logging.info("Ensemble evaluation [%d/%s]", evals_completed,
                       self.steps or "??")
        sess.run(objective_metrics)
      except tf.errors.OutOfRangeError:
        logging.info("Encountered end of input after %d evaluations",
                     evals_completed)
        break

    # Evaluating the first element is idempotent for metric tuples.
    return sess.run([metric[0] for metric in objective_metrics])