tf.contrib.checkpoint.Checkpointable

Manages dependencies on other objects.

Inherits From: CheckpointableBase

Trackable objects may have dependencies: other Trackable objects which should be saved if the object declaring the dependency is saved. A correctly saveable program has a dependency graph such that if changing a global variable affects an object (e.g. changes the behavior of any of its methods) then there is a chain of dependencies from the influenced object to the variable.

Dependency edges have names, and are created implicitly when a Trackable object is assigned to an attribute of another Trackable object. For example:

obj = Trackable()
obj.v = ResourceVariable(0.)

The Trackable object obj now has a dependency named "v" on a variable.

Trackable objects may specify Tensors to be saved and restored directly (e.g. a Variable indicating how to save itself) rather than through dependencies on other objects. See Trackable._gather_saveables_for_checkpoint for details.

© 2020 The TensorFlow Authors. All rights reserved.
Licensed under the Creative Commons Attribution License 3.0.
Code samples licensed under the Apache 2.0 License.
https://www.tensorflow.org/versions/r1.15/api_docs/python/tf/contrib/checkpoint/Checkpointable