tf.types.experimental.distributed.Mirrored

Holds a distributed value: a map from replica id to synchronized values.

Inherits From: DistributedValues

Mirrored values are tf.distribute.DistributedValues for which we know that the value on all replicas is the same. Mirrored values are kept synchronized by the distribution strategy in use, while tf.types.experimental.PerReplica values are left unsynchronized. Mirrored values typically represent model weights. We can safely read a Mirrored value in a cross-replica context by using the value on any replica, while PerReplica values should not be read or manipulated directly by the user in a cross-replica context.

Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License, and code samples are licensed under the Apache 2.0 License. For details, see the Google Developers Site Policies. Java is a registered trademark of Oracle and/or its affiliates. Some content is licensed under the numpy license.

Last updated 2023-03-17 UTC.

English
中文 – 简体

tf.types.experimental.distributed.Mirrored Stay organized with collections Save and categorize content based on your preferences.

tf.types.experimental.distributed.Mirrored