Attend the Women in ML Symposium on December 7 Register now


Stay organized with collections Save and categorize content based on your preferences.

Creates a ClientWorkProcess for federated averaging.

This client work is constructed in slightly different manners depending on whether optimizer is a tff.learning.optimizers.Optimizer, or a no-arg callable returning a tf.keras.optimizers.Optimizer.

If it is a tff.learning.optimizers.Optimizer, we avoid creating tf.Variables associated with the optimizer state within the scope of the client work, as they are not necessary. This also means that the client's model weights are updated by computing and then assigning the result to the model weights (while a tf.keras.optimizers.Optimizer will modify the model weight in place using optimizer.apply_gradients).

model_fn A no-arg function that returns a tff.learning.Model. This method must not capture TensorFlow tensors or variables and use them. The model must be constructed entirely from scratch on each invocation, returning the same pre-constructed model each call will result in an error.
optimizer A tff.learning.optimizers.Optimizer, or a no-arg callable that returns a tf.keras.Optimizer. If using a tf.keras.Optimizer, the resulting process will have no hyperparameters in its state (ie. process.get_hparams will return an empty dictionary), while if using a tff.learning.optimizers.Optimizer, the process will have the same hyperparameters as the optimizer.
client_weighting A tff.learning.ClientWeighting value.
metrics_aggregator A function that takes in the metric finalizers (i.e., tff.learning.Model.metric_finalizers()) and a tff.types.StructWithPythonType of the unfinalized metrics (i.e., the TFF type of tff.learning.Model.report_local_unfinalized_metrics()), and returns a tff.Computation for aggregating the unfinalized metrics. If None, this is set to tff.learning.metrics.sum_then_finalize.
use_experimental_simulation_loop Controls the reduce loop function for input dataset. An experimental reduce loop is used for simulation. It is currently necessary to set this flag to True for performant GPU simulations.

A ClientWorkProcess.