Cannot import keras after installation

Question

I'm trying to setup keras deep learning library for Python3.5 on Ubuntu 16.04 LTS and use Tensorflow as a backend. I have Python2.7 and Python3.5 installed. I have installed Anaconda and with help of it Tensorflow, numpy, scipy, pyyaml. Afterwards I have installed keras with command

sudo python setup.py install

Although I can see /usr/local/lib/python3.5/dist-packages/Keras-1.1.0-py3.5.egg directory, I cannot use keras library. When I try to import it in python it says

ImportError: No module named 'keras'

I have tried to install keras usingpip3, but got the same result.

What am I doing wrong? Any Ideas?

Answer 1

Ran to the same issue, Assuming your using anaconda3 and your using a venv with >= python=3.6:

python -m pip install keras sudo python -m pip install --user tensorflow 

Answer 2

Diagnose

If you have pip installed (you should have it until you use Python 3.5), list the installed Python packages, like this:

$ pip list | grep -i keras Keras (1.1.0) 

If you don’t see Keras, it means that the previous installation failed or is incomplete (this lib has this dependancies: numpy (1.11.2), PyYAML (3.12), scipy (0.18.1), six (1.10.0), and Theano (0.8.2).)

Consult the pip.log to see what’s wrong.

You can also display your Python path like this:

$ python3 -c 'import sys, pprint; pprint.pprint(sys.path)' ['',  '/Library/Frameworks/Python.framework/Versions/3.5/lib/python35.zip',  '/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5',  '/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/plat-darwin',  '/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/lib-dynload',  '/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages'] 

Make sure the Keras library appears in the /Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages path (the path is different on Ubuntu).

If not, try do uninstall it, and retry installation:

$ pip uninstall Keras 

Use a virtualenv

It’s a bad idea to use and pollute your system-wide Python. I recommend using a virtualenv (see this guide).

The best usage is to create a virtualenv directory (in your home, for instance), and store your virtualenvs in:

cd virtualenv/ virtualenv -p python3.5 py-keras source py-keras/bin/activate pip install -q -U pip setuptools wheel 

Then install Keras:

pip install keras 

You get:

$ pip list Keras (1.1.0) numpy (1.11.2) pip (8.1.2) PyYAML (3.12) scipy (0.18.1) setuptools (28.3.0) six (1.10.0) Theano (0.8.2) wheel (0.30.0a0) 

But, you also need to install extra libraries, like Tensorflow:

$ python -c "import keras" Using TensorFlow backend. Traceback (most recent call last):   ... ImportError: No module named 'tensorflow' 

The installation guide of TesnsorFlow is here: https://www.tensorflow.org/versions/r0.11/get_started/os_setup.html#pip-installation