Module: tf.autograph

Conversion of plain Python into TensorFlow graph code.

Note: In TensorFlow 2.0, AutoGraph is automatically applied when using tf.function. This module contains lower-level APIs for advanced use.

For more information, see the AutoGraph guide.

By equivalent graph code we mean code that generates a TensorFlow graph when run. The generated graph has the same effects as the original code when executed (for example with tf.function or tf.compat.v1.Session.run). In other words, using AutoGraph can be thought of as running Python in TensorFlow.

Modules

experimental module: Public API for tf.autograph.experimental namespace.

Functions

set_verbosity(...): Sets the AutoGraph verbosity level.

to_code(...): Similar to to_graph, but returns Python source code as a string.

to_graph(...): Converts a Python entity into a TensorFlow graph.

trace(...): Traces argument information at compilation time.