AttributeError: module 'tensorflow.python.training.experimental.mixed_precision' has no attribute '_register_wrapper_optimizer_cls'

Question

I am using Keras to implement a neural network. But when I use model = Sequential(), I get the following error:

AttributeError                            Traceback (most recent call last)
<ipython-input-31-fa9fd3b0e211> in <module>
      8 
      9 # Create the model
---> 10 model = Sequential()
     11 model.add(Dropout(0.1), input_shape=(128,))
     12 model.add(Dense(256, activation='relu', kernel_initializer='he_uniform'))

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/utils/version_utils.py in __new__(cls, *args, **kwargs)
     55     use_v2 = should_use_v2()
     56     cls = swap_class(cls, training.Model, training_v1.Model, use_v2)  # pylint: disable=self-cls-assignment
---> 57     return super(ModelVersionSelector, cls).__new__(cls)
     58 
     59 

/usr/local/lib/python3.6/dist-packages/tensorflow/python/util/lazy_loader.py in __getattr__(self, item)
     60 
     61   def __getattr__(self, item):
---> 62     module = self._load()
     63     return getattr(module, item)
     64 

/usr/local/lib/python3.6/dist-packages/tensorflow/python/util/lazy_loader.py in _load(self)
     43     """Load the module and insert it into the parent's globals."""
     44     # Import the target module and insert it into the parent's namespace
---> 45     module = importlib.import_module(self.__name__)
     46     self._parent_module_globals[self._local_name] = module
     47 

/usr/lib/python3.6/importlib/__init__.py in import_module(name, package)
    124                 break
    125             level += 1
--> 126     return _bootstrap._gcd_import(name[level:], package, level)
    127 
    128 

/usr/lib/python3.6/importlib/_bootstrap.py in _gcd_import(name, package, level)

/usr/lib/python3.6/importlib/_bootstrap.py in _find_and_load(name, import_)

/usr/lib/python3.6/importlib/_bootstrap.py in _find_and_load_unlocked(name, import_)

/usr/lib/python3.6/importlib/_bootstrap.py in _load_unlocked(spec)

/usr/lib/python3.6/importlib/_bootstrap_external.py in exec_module(self, module)

/usr/lib/python3.6/importlib/_bootstrap.py in _call_with_frames_removed(f, *args, **kwds)

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/base_layer_v1.py in <module>
     48 from tensorflow.python.keras.engine import input_spec
     49 from tensorflow.python.keras.mixed_precision import autocast_variable
---> 50 from tensorflow.python.keras.mixed_precision import loss_scale_optimizer
     51 from tensorflow.python.keras.mixed_precision import policy
     52 from tensorflow.python.keras.saving.saved_model import layer_serialization

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/mixed_precision/loss_scale_optimizer.py in <module>
   1152 
   1153 # pylint: disable=protected-access
-> 1154 mixed_precision._register_wrapper_optimizer_cls(optimizer_v2.OptimizerV2,
   1155                                                 LossScaleOptimizerV1)
   1156 

AttributeError: module 'tensorflow.python.training.experimental.mixed_precision' has no attribute '_register_wrapper_optimizer_cls'

I am relatively new to deep learning. Any help would be appreciated. Thank you!

Answer 1

This problem maybe occurs in the keras package version.

Because the installed tensorflow version does not match the keras version. Just uninstall keras and reinstall your own tensorflow corresponding version.

pip3 uninstall keras
pip3 install keras --upgrade

In summary, you need to match the correct environment version for your Keras code.

Answer 2

I had the same problem. The problem is that Keras is not installed compatible with Tensorflow, so there are no attributes/methods in it. I tried the following commands mentioned in the source and my problem was completely solved.

pip3 uninstall keras
pip3 install keras --upgrade

Answer 3

This is a terrible solution, but for the sake of getting things going, here is how I solved it.

I went into my directory and found the location of tensorflow.python.training.experimental.mixed_precision.py, and manually inserted the following lines

def _register_wrapper_optimizer_cls(optimizer_cls, wrapper_optimizer_cls):
  _REGISTERED_WRAPPER_OPTIMIZER_CLS[optimizer_cls] = wrapper_optimizer_cls

The code is from the official GitHub https://github.com/sourcecode369/tensorflow-1/blob/master/tensorflow/python/training/experimental/mixed_precision.py