org.tensorflow

Class EagerSession.Options

java.lang.Object

org.tensorflow.EagerSession.Options

Enclosing class:

EagerSession

public static class EagerSession.Options
extends Object

Method Summary

Modifier and Type	Method and Description
EagerSession.Options	async(boolean value) Controls how operations dispatched are actually executed.
EagerSession	build() Builds an eager session with the selected options.
EagerSession.Options	config(byte[] value) Configures the session based on the data found in the provided buffer, which is serialized TensorFlow config proto.
EagerSession.Options	devicePlacementPolicy(EagerSession.DevicePlacementPolicy value) Controls how to act when we try to run an operation on a given device but some input tensors are not on that device.
EagerSession.Options	resourceCleanupStrategy(EagerSession.ResourceCleanupStrategy value) Controls how TensorFlow resources are cleaned up when no longer needed.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Detail

async

public EagerSession.Options async(boolean value)

Controls how operations dispatched are actually executed.

When set to true, each operation are executed asynchronously (in which case some operations might return "non-ready" outputs). When set to false, all operations are executed synchronously.

Synchronous execution is used by default.

Parameters:

value - true for asynchronous execution, false for synchronous.

devicePlacementPolicy

public EagerSession.Options devicePlacementPolicy(EagerSession.DevicePlacementPolicy value)

Controls how to act when we try to run an operation on a given device but some input tensors are not on that device.

EagerSession.DevicePlacementPolicy.SILENT is used by default.

Parameters:

value - policy to apply

See Also:

EagerSession.DevicePlacementPolicy

resourceCleanupStrategy

public EagerSession.Options resourceCleanupStrategy(EagerSession.ResourceCleanupStrategy value)

Controls how TensorFlow resources are cleaned up when no longer needed.

EagerSession.ResourceCleanupStrategy.IN_BACKGROUND is used by default.

Parameters:

value - strategy to use

See Also:

EagerSession.ResourceCleanupStrategy

config

public EagerSession.Options config(byte[] value)

Configures the session based on the data found in the provided buffer, which is serialized TensorFlow config proto.

Warning: the support of this feature is subject to changes since TensorFlow protos might not be supported on public endpoints in the future.

Parameters:

value - a serialized config proto

See Also:

build

public EagerSession build()

Builds an eager session with the selected options.