Datasets: TF 2.0 AttributeError: module 'tensorflow._api.v2.autograph.experimental' has no attribute 'do_not_convert'

[NMVRodrigues]

Short description
Its not possible to import tensorflow_datasets using Tensorflow 2.0.
This error appears every time: AttributeError: module 'tensorflow._api.v2.autograph.experimental' has no attribute 'do_not_convert'

Environment information

• Operating System: win10 and Ubunto 18
• Python version: 3.6.8
• tensorflow-datasets/tfds-nightly version: tensorflow-datasets-1.1.0
• tensorflow/tensorflow-gpu/tf-nightly/tf-nightly-gpu version: Tensorflow 2.0.0-alpha0

import tensorflow_datasets as tfds

>>> import tensorflow_datasets as tfds
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "C:\Users\Utilizador\Miniconda3\envs\venv\lib\site-packages\tensorflow_datasets\__init__.py", line 46, in <module>
    from tensorflow_datasets.core import tf_compat
  File "C:\Users\Utilizador\Miniconda3\envs\venv\lib\site-packages\tensorflow_datasets\core\__init__.py", line 28, in <module>
    from tensorflow_datasets.core.dataset_info import DatasetInfo
  File "C:\Users\Utilizador\Miniconda3\envs\venv\lib\site-packages\tensorflow_datasets\core\dataset_info.py", line 53, in <module>
    from tensorflow_datasets.core.features import top_level_feature
  File "C:\Users\Utilizador\Miniconda3\envs\venv\lib\site-packages\tensorflow_datasets\core\features\__init__.py", line 27, in <module>
    from tensorflow_datasets.core.features.features_dict import FeaturesDict
  File "C:\Users\Utilizador\Miniconda3\envs\venv\lib\site-packages\tensorflow_datasets\core\features\features_dict.py", line 28, in <module>
    from tensorflow_datasets.core.features import top_level_feature
  File "C:\Users\Utilizador\Miniconda3\envs\venv\lib\site-packages\tensorflow_datasets\core\features\top_level_feature.py", line 27, in <module>
    class TopLevelFeature(feature_lib.FeatureConnector):
  File "C:\Users\Utilizador\Miniconda3\envs\venv\lib\site-packages\tensorflow_datasets\core\features\top_level_feature.py", line 45, in TopLevelFeature
    @tf.autograph.experimental.do_not_convert()
AttributeError: module 'tensorflow._api.v2.autograph.experimental' has no attribute 'do_not_convert'

[us]

Thanks for reporting! Can you check with Tensorflow 2.0.0-beta1 version.

[NMVRodrigues]

Thanks for reporting! Can you check with Tensorflow 2.0.0-beta1 version.

That does solve the issue yes. Much appreciated.

[RahulR432]

Is it possible to import tensorflow_datasets using any other TF 2.x version other than the beta.

[Conchylicultor]

@RahulR432 tf.compat.v2.autograph.experimental.do_not_convert should be available in all version of TF 2.0.0, including the last stable 2.1. https://www.tensorflow.org/api_docs/python/tf/autograph/experimental/do_not_convert