Tensorflow-gpu and Textgenrnn version incompatibility on Windows 10: ModuleNotFoundError: No module named 'tensorflow.contrib'

Question

No matter what I do I cannot seem to get TensorFlow-gpu to work nicely with textgenrnn. The Tensorflow-gpu test is saying everything is running properly Currently, I am getting the error:

ModuleNotFoundError: No module named 'tensorflow.contrib'

when I try to import textgenrnn as

textgen = textgenrnn(name=model_name)

Could someone please provide their working development environment? My pip list is:

Package              Version
-------------------- -------------------
absl-py              0.8.1
astor                0.8.0
bleach               1.5.0
cachetools           3.1.1
certifi              2019.9.11
chardet              3.0.4
gast                 0.2.2
google-auth          1.6.3
google-auth-oauthlib 0.4.1
google-pasta         0.1.7
grpcio               1.24.3
h5py                 2.10.0
html5lib             0.9999999
idna                 2.8
joblib               0.14.0
Keras                2.3.1
Keras-Applications   1.0.8
Keras-Preprocessing  1.1.0
Markdown             3.1.1
numpy                1.17.3
oauthlib             3.1.0
opt-einsum           3.1.0
pip                  19.3.1
protobuf             3.10.0
pyasn1               0.4.7
pyasn1-modules       0.2.7
PyYAML               5.1.2
requests             2.22.0
requests-oauthlib    1.2.0
rsa                  4.0
scikit-learn         0.21.3
scipy                1.3.1
setuptools           41.5.1
six                  1.12.0
tb-nightly           2.1.0a20191029
tensorboard          2.0.1
tensorflow-estimator 2.0.1
tensorflow-gpu       2.0.0
termcolor            1.1.0
textgenrnn           1.5.0
tf-estimator-nightly 2.0.0.dev2019102908
tqdm                 4.36.1
urllib3              1.25.6
Werkzeug             0.16.0
wheel                0.33.6
wrapt                1.11.2

And my Python Version is 3.6.4, CUDA 10 and cudnn 7.4.1.5. I have tried with lower/higher python versions and CUDA 9 as well. Thanks.

Answer 1

Speak of the heckin devil, I finally got it after a few days, albeit with console warnings.

Ended up with the above configuration, and uninstalling tensorflow-gpu to install tensorflow-gpu==1.14.0.

I hope this can help someone.

Answer 2

It seems to be that there is a problem with tensorflow version. As a contrib module, this is no longer available in TensorFlow 2.0 How about you install again your tensorflow-gpu with lower version. I prefer v1.14

pip3 instal tensorflow-gpu==1.14