tf.contrib.eager

TensorFlow Eager execution prototype.

EXPERIMENTAL: APIs here are unstable and likely to change without notice.

To use, at program startup, call tf.compat.v1.enable_eager_execution().

Modules

• metrics module: Metrics namespace.

Classes

• class Checkpoint: Groups trackable objects, saving and restoring them.
• class Checkpointable: Manages dependencies on other objects.
• class EagerVariableStore: Wrapper allowing functional layers to be used with eager execution.
• class ExecutionCallback: Valid callback actions.
• class GradientTape: Record operations for automatic differentiation.
• class Iterator: An iterator producing tf.Tensor objects from a tf.data.Dataset.
• class Network: Represents the composition of a set of Layers.
• class Saver: A tf.compat.v1.train.Saver adapter for use when eager execution is enabled.
• class Sequential: Represents a linear sequence of Layers or functions.
• class TensorSpec: Describes a tf.Tensor.
• class Variable: Variable based on resource handles.

Functions

• add_execution_callback(...): Add an execution callback to the default eager context.
• async_clear_error(...): Clears errors raised during ASYNC execution mode.
• async_wait(...): Waits for ops dispatched in ASYNC mode to finish.
• clear_execution_callbacks(...): Clear all execution callbacks from the default eager context.
• connect_to_remote_host(...): Connects to a single machine to enable remote execution on it.
• custom_gradient(...): Decorator to define a function with a custom gradient.
• defun(...): Compiles a Python function into a callable TensorFlow graph.
• enable_eager_execution(...): Enables eager execution for the lifetime of this program.
• enable_remote_eager_execution(...): Enables eager execution for the lifetime of this program.
• errstate(...): Context manager setting error state.
• executing_eagerly(...): Returns True if the current thread has eager execution enabled.
• execution_mode(...): Context manager for setting execution mode for current thread.
• function(...): Creates a callable TensorFlow graph from a Python function.
• get_optimizer_variables(...): Returns a list of variables for the given tf.compat.v1.train.Optimizer.
• gradients_function(...): Returns a function which differentiates f with respect to params.
• implicit_gradients(...): Returns a function which differentiates f with respect to variables.
• implicit_value_and_gradients(...): Returns a function which differentiates f with respect to variables.
• in_eager_mode(...): Returns True if the current thread has eager execution enabled.
• inf_callback(...): A specialization of inf_nan_callback that checks for infs only.
• inf_nan_callback(...): An execution callback that checks for infs and nans in output tensors.
• list_devices(...): List the names of the available devices.
• make_template(...): Make a template, optionally compiling func_ into a graph function.
• nan_callback(...): A specialization of inf_nan_callback that checks for nans only.
• num_gpus(...): Get the number of available GPU devices.
• py_func(...): Wraps a python function into a TensorFlow op that executes it eagerly.
• restore_network_checkpoint(...): Restore the Network from a checkpoint. (deprecated)
• restore_variables_on_create(...): ContextManager that restores variables on creation.
• run(...): Runs the program with an optional main function and argv list.
• run_all_tests_in_graph_and_eager_modes(...): Execute all test methods in the given class with and without eager.
• run_test_in_graph_and_eager_modes(...): Execute the decorated test with and without enabling eager execution.
• save_network_checkpoint(...): Save variables from the Network to a checkpoint. (deprecated)
• set_execution_mode(...): Sets execution mode for the current thread.
• set_server_def(...)
• seterr(...): Set how abnormal conditions are handled by the default eager context.
• value_and_gradients_function(...): Returns a function that computes f and its derivative w.r.t. params.