Tools for working with object-based checkpoints.
Visualization and inspection:
Trackable data structures:
Saving and restoring Python state:
class CheckpointManager: Deletes old checkpoints.
class Checkpointable: Manages dependencies on other objects.
class CheckpointableBase: Base class for
Trackable objects without automatic dependencies.
class List: An append-only sequence type which is trackable.
class Mapping: An append-only trackable mapping data structure with string keys.
class NoDependency: Allows attribute assignment to
Trackable objects with no dependency.
class NumpyState: A trackable object whose NumPy array attributes are saved/restored.
class PythonStateWrapper: A mixin for putting Python state in an object-based checkpoint.
class UniqueNameTracker: Adds dependencies on trackable objects with name hints.
capture_dependencies(...): Capture variables created within this scope as
list_objects(...): Traverse the object graph and list all accessible objects.
object_metadata(...): Retrieves information about the objects in a checkpoint.
split_dependency(...): Creates multiple dependencies with a synchronized save/restore.