Conversion of eager-style Python into TensorFlow graph code.
AutoGraph transforms a subset of Python which operates on TensorFlow objects
into equivalent TensorFlow graph code. When executing the graph, it has the same
effect as if you ran the original code in eager mode.
Python code which doesn't operate on TensorFlow objects remains functionally
unchanged, but keep in mind that tf.function
only executes such code at trace
time, and generally will not be consistent with eager execution.
For more information, see the AutoGraph reference documentation, and the tf.function guide.
Modules
experimental
module: Public API for tf.autograph.experimental namespace.
Functions
set_verbosity(...)
: Sets the AutoGraph verbosity level.
to_code(...)
: Returns the source code generated by AutoGraph, as a string.
to_graph(...)
: Converts a Python entity into a TensorFlow graph.
trace(...)
: Traces argument information at compilation time.