tff.simulation.compose_dataset_computation_with_iterative_process  |  TensorFlow Federated (original) (raw)

tff.simulation.compose_dataset_computation_with_iterative_process

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

Builds a new iterative process which constructs datasets on clients.

tff.simulation.compose_dataset_computation_with_iterative_process(
    dataset_computation: tff.Computation,
    process: tff.templates.IterativeProcess
) -> tff.templates.IterativeProcess

Used in the notebooks

Given a tff.Computation that returns a tf.data.Dataset, and atff.templates.IterativeProcess where exactly one of the arguments is a dataset placed on clients of the same type as returned by thetff.Computation, this function will construct a newtff.templates.IterativeProcess whose next function accepts a federated set of values of the same type as the parameter of the dataset_computation, mapsdataset_computation over these values, and proceeds with the body ofprocess.next.

For example, if the type signature of dataset_computation is:

(T -> U*)

and the type signature of process.next is:

(<S, {U*}@CLIENTS> -> <S, V>

then the returned tff.templates.IterativeProcess.next type signature will be:

(<S, {T}@CLIENTS> -> <S, V>)

This functionality is useful in several settings:

• We may want to push some dataset preprocessing to happen on the clients, as opposed to preprocessing happening on the TFF simultation controller. This may be necessary, e.g., in the case that we want to shuffle client examples.
• We may want to construct the entire dataset on the clients, given a client id. This may be useful in order to speed up distributed simulations, in order to remove a linear cost incurred in constructing and serializign the datasets on the controller.