Collectives™ on Stack Overflow

Find centralized, trusted content and collaborate around the technologies you use most.

Teams

Q&A for work

Connect and share knowledge within a single location that is structured and easy to search.

raises the following error:

AttributeError: module 'tensorflow' has no attribute 'name_scope'.
I am using the following packages.
Keras                     2.2.2,                     
Keras-Applications        1.0.4,                   
Keras-Preprocessing       1.0.2,                   
tensorflow                1.9.0,                     
tensorflow-gpu            1.9.0                
I was unable to reproduce with the same versions of the keras and tensorflow, reinstalling keras and tensorflow, may solve the issue, please use commands below:
pip install --upgrade pip setuptools wheel
pip install -I tensorflow
pip install -I keras
NOTE: The -I parameter stands for ignore installed package.
                Just to add to this answer, name_scope should indeed be included with tf 1.9.0 so it seems that there is something fishy with the installation.
– VegardKT
                Aug 7, 2018 at 10:29
                But Why? I was using this environment for a week now to build models in keras. Maybe because I installed tensorflow serving today?
– Henrik
                Aug 7, 2018 at 10:42
                I get the error "ImportError: Keras requires TensorFlow 2.2 or higher. Install TensorFlow via pip install tensorflow" when following these instructions.
– user3367130
                Aug 27, 2020 at 2:10
                @user3367130 I think you have installed older tensorflow version at some point, you can check versions with pip freeze
– Andriy Ivaneyko
                Aug 27, 2020 at 9:02
As Andriy Ivaneyko mentioned above, reinstalling tensorflow helps. I'm not sure why, but installing tensorflow-serving-api breaks something somewhere along the way. We solved this by running:
pip install --force-reinstall tensorflow
Note that this applies to both tensorflow and tensorflow-gpu installations. Specifically, the above command will fix this problem in situations where you're specifically using tensorlfow-gpu. tensorflow-serving-api installs regular tensorflow if it's not already installed.
I encountered this same bug and reinstalling tensorflow made no difference, and this caused me some headscratching. 
Eventually I noticed that my IDE autocomplete had added the following line to my code:
from tensorflow_core.python.keras.callbacks import EarlyStopping
It seems that directly referencing tensorflow_core.python will break tensorflow.  
Replacing this with the normal tensorflow import solved the problem! 
from tensorflow.keras.callbacks import EarlyStopping
In my example I could either import like this:
from keras.layers import Dense, Dropout, LSTM, Input, Activation
from keras import optimizers, Model
or like that:
from tensorflow_core.python.keras import Input, Model, optimizers
from tensorflow_core.python.keras.layers import LSTM, Dropout, Dense
Mixing up tensorflow_core.python.keras and plain keras led to the problem in my case. After I imported everything directly from keras and keras.layers, it worked for me.
        
Thanks for contributing an answer to Stack Overflow!
Please be sure to answer the question. Provide details and share your research!
But avoid …
Asking for help, clarification, or responding to other answers.
Making statements based on opinion; back them up with references or personal experience.
To learn more, see our tips on writing great answers.