TensorFlow 2.0 Beta is available Learn more

tff.learning.state_with_new_model_weights

Returns a ServerState with updated model weights.

tff.learning.state_with_new_model_weights(
    server_state,
    trainable_weights,
    non_trainable_weights
)

Defined in python/learning/framework/optimizer_utils.py.

Args:

  • server_state: A server state object returned by an iterative training process like tff.learning.build_federated_averaging_process.
  • trainable_weights: A list of numpy arrays in the order of the original model's trainable_variables.
  • non_trainable_weights: A list of numpy arrays in the order of the original model's non_trainable_variables.

Returns:

A new server ServerState object which can be passed to the next method of the iterative process.