Watch talks from the 2019 TensorFlow Dev Summit Watch now

tf.contrib.checkpoint.object_metadata

tf.contrib.checkpoint.object_metadata(save_path)

Defined in tensorflow/python/training/checkpointable/util.py.

Retrieves information about the objects in a checkpoint.

Example usage:

object_graph = tf.contrib.checkpoint.object_metadata(
    tf.train.latest_checkpoint(checkpoint_directory))
ckpt_variable_names = set()
for node in object_graph.nodes:
  for attribute in node.attributes:
    ckpt_variable_names.add(attribute.full_name)

Args:

Returns:

A parsed tf.contrib.checkpoint.CheckpointableObjectGraph protocol buffer.

Raises:

  • ValueError: If an object graph was not found in the checkpoint.