tf.compat.v1.flags.tf_decorator.rewrap
Injects a new target into a function built by make_decorator.
tf.compat.v1.flags.tf_decorator.rewrap( decorator_func, previous_target, new_target )
This function allows replacing a function wrapped by decorator_func
, assuming the decorator that wraps the function is written as described below.
The decorator function must use <decorator name>.__wrapped__
instead of the wrapped function that is normally used:
Example:
Instead of this:
def simple_parametrized_wrapper(*args, *kwds): return wrapped_fn(args, **kwds)
tf_decorator.make_decorator(simple_parametrized_wrapper, wrapped_fn)
Write this:
def simple_parametrized_wrapper(*args, *kwds): return simple_parametrizedwrapper.wrapped_(args, **kwds)
tf_decorator.make_decorator(simple_parametrized_wrapper, wrapped_fn)
Note that this process modifies decorator_func.
Args | |
---|---|
decorator_func | Callable returned by wrap . |
previous_target | Callable that needs to be replaced. |
new_target | Callable to replace previous_target with. |
Returns | |
---|---|
The updated decorator. If decorator_func is not a tf_decorator, new_target is returned. |
© 2020 The TensorFlow Authors. All rights reserved.
Licensed under the Creative Commons Attribution License 3.0.
Code samples licensed under the Apache 2.0 License.
https://www.tensorflow.org/versions/r2.4/api_docs/python/tf/compat/v1/flags/tf_decorator/rewrap