Support for training models.
See the Training guide.
Modules
experimental module: Public API for tf.train.experimental namespace.
Classes
class BytesList: Used in tf.train.Example protos. Holds a list of byte-strings.
class Checkpoint: Manages saving/restoring trackable values to disk.
class CheckpointManager: Manages multiple checkpoints by keeping some and deleting unneeded ones.
class CheckpointOptions: Options for constructing a Checkpoint.
class ClusterDef: A ProtocolMessage
class ClusterSpec: Represents a cluster as a set of "tasks", organized into "jobs".
class Coordinator: A coordinator for threads.
class Example: An Example is a standard proto storing data for training and inference.
class ExponentialMovingAverage: Maintains moving averages of variables by employing an exponential decay.
class Feature: Used in tf.train.Example protos. Contains a list of values.
class FeatureList: Mainly used as part of a tf.train.SequenceExample.
class FeatureLists: Mainly used as part of a tf.train.SequenceExample.
class Features: Used in tf.train.Example protos. Contains the mapping from keys to Feature.
class FloatList: Used in tf.train.Example protos. Holds a list of floats.
class Int64List: Used in tf.train.Example protos. Holds a list of Int64s.
class JobDef: A ProtocolMessage
class SequenceExample: A SequenceExample represents a sequence of features and some context.
class ServerDef: A ProtocolMessage
class TrackableView: Gathers and serializes a trackable view.
Functions
checkpoints_iterator(...): Continuously yield new checkpoint files as they appear.
get_checkpoint_state(...): Returns CheckpointState proto from the "checkpoint" file.
latest_checkpoint(...): Finds the filename of latest saved checkpoint file.
list_variables(...): Lists the checkpoint keys and shapes of variables in a checkpoint.
load_checkpoint(...): Returns CheckpointReader for checkpoint found in ckpt_dir_or_file.
load_variable(...): Returns the tensor value of the given variable in the checkpoint.