tff.framework.LambdaExecutor

View source on GitHub

The lambda executor handles lambda expressions and related abstractions.

Inherits From: Executor

tff.framework.LambdaExecutor(
    target_executor
)

This executor understands TFF computation compositional constructs, including lambdas, blocks, references, calls, tuples, and selections, and orchestrates the execution of these constructs, while delegating all the non-compositional constructs (tensorflow, intrinsics, data, or placement) to a target executor.

NOTE: Not all lambda expressions are executed by this lambda executor. If the computation contains a call to an instrinsic that takes a functional argument, that functional argument is fed in its entirety to the target executor rather than being parsed by the lambda executor (since its execution needs to happen elsewhere).

The arguments to be ingested can be either federated computations (those are natively interpreted), or whatever other form of arguments are understood by the target executor.

Args:

  • target_executor: An instance of executor_base.Executor to which the lambda executor delegates all that it cannot handle by itself.

Methods

close

View source

close()

Release resources associated with this Executor, if any.

If the executor has one or more target Executors, implementation of this method must close them.

create_call

View source

create_call(
    *args, **kwargs
)

create_selection

View source

create_selection(
    *args, **kwargs
)

create_tuple

View source

create_tuple(
    *args, **kwargs
)

create_value

View source

create_value(
    *args, **kwargs
)