def main(argv=None):
dataset = _get_input()
# Extract input tensors for evaluation
iterator = tf.compat.v1.data.make_one_shot_iterator(dataset)
features, labels = iterator.get_next()
# Construct the evaluation function
evaluate_fn = model_fn.evaluate_fn()
# Wrap the ops in an Estimator spec object
estimator_spec = evaluate_fn(features, labels, tf.estimator.ModeKeys.EVAL,
{'continuous_eval': True})
# Extract the necessary ops and the final tensors from the estimator spec
update_op, value_ops = _extract_metric_update_ops(
estimator_spec.eval_metric_ops)
# Specify to evaluate N number of batches (in this case N==1)
stop_hook = evaluation._StopAfterNEvalsHook(1)
# Create summaries of values added to tf.GraphKeys.SUMMARIES
summary_writer = tf.compat.v1.summary.FileWriter(
os.path.join(FLAGS.model, FLAGS.output))
summary_hook = evaluation.SummaryAtEndHook(summary_writer=summary_writer)
# Evaluate repeatedly once a new checkpoint is found
evaluation.evaluate_repeatedly(checkpoint_dir=FLAGS.model,
eval_ops=update_op,
final_ops=value_ops,
hooks=[stop_hook, summary_hook],
config=_get_config(),
eval_interval_secs=FLAGS.eval_interval_secs)
def main(argv=None):
# Initialize the classifier
classifier = tf.estimator.Estimator(config=_get_config(),
model_fn=model_fn.evaluate_fn(),
model_dir=FLAGS.model,
params={'continuous_eval': False})
evaluations = classifier.evaluate(input_fn=_get_input)
print(evaluations)