diff options
author | XFeiF <eva.aeolus@gmail.com> | 2018-07-21 22:59:03 +0800 |
---|---|---|
committer | XFeiF <eva.aeolus@gmail.com> | 2018-07-21 22:59:03 +0800 |
commit | 828257c82a6dfc1537547d226b25b7e394ff3cd4 (patch) | |
tree | 1f2a935928838f6d7c88fd152814bef4f2eba08e /tensorflow/contrib/slim | |
parent | dae7a75734f2137aae7130e064fab9dfcb799c45 (diff) |
[tf.contrib.slim] Update documentation in evaluation.py
Diffstat (limited to 'tensorflow/contrib/slim')
-rw-r--r-- | tensorflow/contrib/slim/python/slim/evaluation.py | 25 |
1 files changed, 15 insertions, 10 deletions
diff --git a/tensorflow/contrib/slim/python/slim/evaluation.py b/tensorflow/contrib/slim/python/slim/evaluation.py index 5cfd5ee82e..0feb3925eb 100644 --- a/tensorflow/contrib/slim/python/slim/evaluation.py +++ b/tensorflow/contrib/slim/python/slim/evaluation.py @@ -22,7 +22,8 @@ modules using a variety of metrics and summarizing the results. ********************** In the simplest use case, we use a model to create the predictions, then specify -the metrics and finally call the `evaluation` method: +the metrics and choose one model checkpoint, finally call the`evaluation_once` +method: # Create model and obtain the predictions: images, labels = LoadData(...) @@ -34,20 +35,24 @@ the metrics and finally call the `evaluation` method: "mse": slim.metrics.mean_squared_error(predictions, labels), }) + checkpoint_path = '/tmp/my_model_dir/my_checkpoint' + log_dir = '/tmp/my_model_eval/' + initial_op = tf.group( tf.global_variables_initializer(), tf.local_variables_initializer()) - with tf.Session() as sess: - metric_values = slim.evaluation( - sess, - num_evals=1, - initial_op=initial_op, - eval_op=names_to_updates.values(), - final_op=name_to_values.values()) + metric_values = slim.evaluate_once( + master='', + checkpoint_path=checkpoint_path, + log_dir=log_dir, + num_evals=1, + initial_op=initial_op, + eval_op=names_to_updates.values(), + final_op=name_to_values.values()) - for metric, value in zip(names_to_values.keys(), metric_values): - logging.info('Metric %s has value: %f', metric, value) + for metric, value in zip(names_to_values.keys(), metric_values): + logging.info('Metric %s has value: %f', metric, value) ************************************************ * Evaluating a Checkpointed Model with Metrics * |