View source on GitHub
|
Libraries for developing federated learning algorithms.
Classes
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.
Functions
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.