Keras using theano backend despite keras.json specifying tensorflow

Question

Keras is using theano as a backend despite the .keras/keras.json file specifying tensorflow.

(conda_env) [centos@ip ~]$ cat .keras/keras.json
{
    "epsilon": 1e-07, 
    "floatx": "float32", 
    "image_data_format": "channels_first", 
    "backend": "tensorflow"
}

Python 2.7.15 | packaged by conda-forge | (default, Nov 28 2018, 18:42:13) 
[GCC 7.3.0] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import keras
Using Theano backend.

I have tensorflow installed and can import it without a problem. Where is keras getting the command to use theano as a backend and how can I change it?

Answer 1

The issue is with anaconda, which is using its own file to determine the backend. You'll need to edit that file, which should be located somewhere like /home/user/anaconda/envs/my_env/etc/conda/activate.d/keras_activate.sh