Module: tff.learning.framework

Libraries for developing federated learning algorithms.


class ClientDeltaFn: Represents a client computation that produces an update to a model.

class ClientOutput: Structure for outputs returned from clients during federated optimization.

class ModelWeights: A container for the trainable and non-trainable variables of a Model.

class ServerState: Represents the state of the server carried between rounds.


build_encoded_broadcast_process_from_model(...): Builds MeasuredProcess for weights of model returned by model_fn.

build_model_delta_optimizer_process(...): Constructs tff.templates.IterativeProcess for Federated Averaging or SGD.

build_stateless_broadcaster(...): Builds a MeasuredProcess that wraps tff.federated_broadcast.

weights_type_from_model(...): Creates a tff.Type from a tff.learning.Model or callable that constructs a model.