Keras: module keras not found after successful installation

[pbhogale]

I managed to install keras with

install.packages("reticulate")
install.packages("keras")
library(reticulate)
library(keras)
install_keras(method = "conda", conda = "/home/prasanna/anaconda2/envs/r-tensorflow/bin/conda", tensorflow = "gpu")

giving the output

Using r-tensorflow conda environment for TensorFlow installation
Determining latest release of TensorFlow...done
Installing TensorFlow...
Collecting tensorflow-gpu==1.3.0 from https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.3.0-cp27-none-linux_x86_64.whl
  Using cached https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.3.0-cp27-none-linux_x86_64.whl
Collecting six>=1.10.0 (from tensorflow-gpu==1.3.0)
  Using cached six-1.10.0-py2.py3-none-any.whl
Collecting protobuf>=3.3.0 (from tensorflow-gpu==1.3.0)
  Using cached protobuf-3.4.0-cp27-cp27mu-manylinux1_x86_64.whl
Collecting wheel (from tensorflow-gpu==1.3.0)
  Using cached wheel-0.29.0-py2.py3-none-any.whl
Collecting backports.weakref>=1.0rc1 (from tensorflow-gpu==1.3.0)
  Using cached backports.weakref-1.0rc1-py2-none-any.whl
Collecting numpy>=1.11.0 (from tensorflow-gpu==1.3.0)
  Using cached numpy-1.13.1-cp27-cp27mu-manylinux1_x86_64.whl
Collecting tensorflow-tensorboard<0.2.0,>=0.1.0 (from tensorflow-gpu==1.3.0)
  Using cached tensorflow_tensorboard-0.1.6-py2-none-any.whl
Collecting mock>=2.0.0 (from tensorflow-gpu==1.3.0)
  Using cached mock-2.0.0-py2.py3-none-any.whl
Collecting setuptools (from protobuf>=3.3.0->tensorflow-gpu==1.3.0)
  Using cached setuptools-36.4.0-py2.py3-none-any.whl
Collecting werkzeug>=0.11.10 (from tensorflow-tensorboard<0.2.0,>=0.1.0->tensorflow-gpu==1.3.0)
  Using cached Werkzeug-0.12.2-py2.py3-none-any.whl
Collecting html5lib==0.9999999 (from tensorflow-tensorboard<0.2.0,>=0.1.0->tensorflow-gpu==1.3.0)
Collecting markdown>=2.6.8 (from tensorflow-tensorboard<0.2.0,>=0.1.0->tensorflow-gpu==1.3.0)
Collecting bleach==1.5.0 (from tensorflow-tensorboard<0.2.0,>=0.1.0->tensorflow-gpu==1.3.0)
  Using cached bleach-1.5.0-py2.py3-none-any.whl
Collecting funcsigs>=1; python_version < "3.3" (from mock>=2.0.0->tensorflow-gpu==1.3.0)
  Using cached funcsigs-1.0.2-py2.py3-none-any.whl
Collecting pbr>=0.11 (from mock>=2.0.0->tensorflow-gpu==1.3.0)
  Using cached pbr-3.1.1-py2.py3-none-any.whl
Installing collected packages: six, setuptools, protobuf, wheel, backports.weakref, numpy, werkzeug, html5lib, markdown, bleach, tensorflow-tensorboard, funcsigs, pbr, mock, tensorflow-gpu
Successfully installed backports.weakref-1.0rc1 bleach-1.5.0 funcsigs-1.0.2 html5lib-0.9999999 markdown-2.6.9 mock-2.0.0 numpy-1.13.1 pbr-3.1.1 protobuf-3.4.0 setuptools-36.4.0 six-1.10.0 tensorflow-gpu-1.3.0 tensorflow-tensorboard-0.1.6 werkzeug-0.12.2 wheel-0.29.0
Fetching package metadata .........
Solving package specifications: .

# All requested packages already installed.
# packages in environment at /home/prasanna/anaconda2/envs/r-tensorflow:
#
h5py                      2.7.0               np113py27_0  
pillow                    4.2.1                    py27_0  
pyyaml                    3.12                     py27_0  
requests                  2.14.2                   py27_0  
scipy                     0.19.1              np113py27_0  

Installation of TensorFlow complete.

Requirement already up-to-date: keras in /home/prasanna/anaconda2/envs/r-tensorflow/lib/python2.7/site-packages
Requirement already up-to-date: pyyaml in /home/prasanna/anaconda2/envs/r-tensorflow/lib/python2.7/site-packages (from keras)
Requirement already up-to-date: numpy>=1.9.1 in /home/prasanna/anaconda2/envs/r-tensorflow/lib/python2.7/site-packages (from keras)
Requirement already up-to-date: six>=1.9.0 in /home/prasanna/anaconda2/envs/r-tensorflow/lib/python2.7/site-packages (from keras)
Requirement already up-to-date: scipy>=0.14 in /home/prasanna/anaconda2/envs/r-tensorflow/lib/python2.7/site-packages (from keras)

Installation of Keras complete.

then I followed that up with

mnist <- dataset_mnist()

which gave the error

Error in py_module_import(module, convert = convert) : ImportError: No module named keras

The same error occurs wven if I install tensorflow without the GPU option.
I looked at old issues similar to this, and could not find a fix. Help would be much appreciated !

[jjallaire]

What is the output of:

library(reticulate)
py_discover_config("keras")

You could try explicitly pointing to the r-tensorflow conda env with:

library(keras)
use_condaenv("r-tensorflow")

If that doesn't work then there may be a problem with locating CUDA / cuDNN (see the GPU installation instructions for details).

[gregmcinnes]

I am having the same issue the original poster described.

Here is what I see when I run py_discover_config. I see a similar output when I try use_condaenv explicitly. Any suggestions?

> library(reticulate)
> py_discover_config("keras")
python:         /home/users/gmcinnes/anaconda3/bin/python
libpython:      /home/users/gmcinnes/anaconda3/lib/libpython3.6m.so
pythonhome:     /home/users/gmcinnes/anaconda3:/home/users/gmcinnes/anaconda3
version:        3.6.4 |Anaconda, Inc.| (default, Jan 16 2018, 18:10:19)  [GCC 7.2.0]
numpy:          /home/users/gmcinnes/anaconda3/lib/python3.6/site-packages/numpy
numpy_version:  1.14.0
keras:          [NOT FOUND]

python versions found:
 /home/users/gmcinnes/anaconda3/bin/python
 /usr/bin/python
 /home/users/gmcinnes/anaconda3_old/bin/python

[theguru111]

@gregmcinnes I run into same issue, did you find the solution for this? Any help will be much appreciated.

[nico-schmid]

Same with me. Since this post is several month old, there should be a fix right?
Is there any solution to this problem?

[nico-schmid]

Hi Everyone,
in case someone does still care...
I was able to fix the problem at least for me. In my case the problem was, that there was a path to an "old" python environment somehow hard coded in my PATH system environmental variables.
You can check on that by eighter using cmdand set PATH or from the R with e.g. strsplit(shell("set path",intern=TRUE),split=";")
After removing the old path and adding the current path to my r-tensorflow environment the system was ready to go. Probably the same should work if you don't add the new path to the system env variables but include it into the ~/.Renviron file. For me, I'm just happy that it works now, so I will not check for better solutions, but if someone does, I'm happy to improve my solution as well ;)