tf.compat.v1.enable_resource_variables

Creates resource variables by default.

tf.compat.v1.enable_resource_variables()

Resource variables are improved versions of TensorFlow variables with a well-defined memory model. Accessing a resource variable reads its value, and all ops which access a specific read value of the variable are guaranteed to see the same value for that tensor. Writes which happen after a read (by having a control or data dependency on the read) are guaranteed not to affect the value of the read tensor, and similarly writes which happen before a read are guaranteed to affect the value. No guarantees are made about unordered read/write pairs.

Calling tf.enable_resource_variables() lets you opt-in to this TensorFlow 2.0 feature.

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.