tff.learning.framework.ServerState

Represents the state of the server carried between rounds.

model A ModelWeights structure, containing Tensors or Variables.
optimizer_state A list of Tensors or Variables, in the order returned by optimizer.variables()
delta_aggregate_state State (possibly empty) of the delta_aggregate_fn.
model_broadcast_state State (possibly empty) of the model_broadcast_fn.