Module: tf.compat

Compatibility functions.

The tf.compat module contains two sets of compatibility functions.

Tensorflow 1.x and 2.x APIs

The compat.v1 and compat.v2 submodules provide a complete copy of both the v1 and v2 APIs for backwards and forwards compatibility across TensorFlow versions 1.x and 2.x. See the migration guide for details.

Utilities for writing compatible code

Aside from the compat.v1 and compat.v2 submodules, tf.compat also contains a set of helper functions for writing code that works in both:

  • TensorFlow 1.x and 2.x
  • Python 2 and 3

Type collections

The compatibility module also provides the following aliases for common sets of python types:

  • bytes_or_text_types
  • complex_types
  • integral_types
  • real_types

Modules

v1 module: Bring in all of the public TensorFlow interface into this module.

Functions

as_bytes(...): Converts bytearray, bytes, or unicode python input types to bytes.

as_str(...)

as_str_any(...): Converts input to str type.

as_text(...): Converts any string-like python input types to unicode.

dimension_at_index(...): Compatibility utility required to allow for both V1 and V2 behavior in TF.

dimension_value(...): Compatibility utility required to allow for both V1 and V2 behavior in TF.

forward_compatibility_horizon(...): Context manager for testing forward compatibility of generated graphs.

forward_compatible(...): Return true if the forward compatibility window has expired.

path_to_str(...): Converts input which is a PathLike object to str type.

bytes_or_text_types (<class 'bytes'>, <class 'str'>)
complex_types (<class 'numbers.Complex'>, <class 'numpy.number'>)
integral_types (<class 'numbers.Integral'>, <class 'numpy.integer'>)
real_types (<class 'numbers.Real'>, <class 'numpy.integer'>, <class 'numpy.floating'>)