Home » Python » Error running basic tensorflow example

Error running basic tensorflow example

Posted by: admin November 30, 2017 Leave a comment

Questions:

I have just reinstalled latest tensorflow on ubuntu:

$ sudo pip install --upgrade https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-0.7.1-cp27-none-linux_x86_64.whl
[sudo] password for ubuntu: 
The directory '/home/ubuntu/.cache/pip/http' or its parent directory is not owned by the current user and the cache has been disabled. Please check the permissions and owner of that directory. If executing pip with sudo, you may want sudo's -H flag.
The directory '/home/ubuntu/.cache/pip' or its parent directory is not owned by the current user and caching wheels has been disabled. check the permissions and owner of that directory. If executing pip with sudo, you may want sudo's -H flag.
Collecting tensorflow==0.7.1 from https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-0.7.1-cp27-none-linux_x86_64.whl
  Downloading https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-0.7.1-cp27-none-linux_x86_64.whl (13.8MB)
    100% |████████████████████████████████| 13.8MB 32kB/s 
Requirement already up-to-date: six>=1.10.0 in /usr/local/lib/python2.7/dist-packages (from tensorflow==0.7.1)
Requirement already up-to-date: protobuf==3.0.0b2 in /usr/local/lib/python2.7/dist-packages (from tensorflow==0.7.1)
Requirement already up-to-date: wheel in /usr/local/lib/python2.7/dist-packages (from tensorflow==0.7.1)
Requirement already up-to-date: numpy>=1.8.2 in /usr/local/lib/python2.7/dist-packages (from tensorflow==0.7.1)
Requirement already up-to-date: setuptools in /usr/local/lib/python2.7/dist-packages (from protobuf==3.0.0b2->tensorflow==0.7.1)
Installing collected packages: tensorflow
  Found existing installation: tensorflow 0.7.1
    Uninstalling tensorflow-0.7.1:
      Successfully uninstalled tensorflow-0.7.1
Successfully installed tensorflow-0.7.1

When following the directions to test it fails with cannot import name pywrap_tensorflow:

$ ipython

/git/tensorflow/tensorflow/__init__.py in <module>()
     21 from __future__ import print_function
     22 
---> 23 from tensorflow.python import *

/git/tensorflow/tensorflow/python/__init__.py in <module>()
     43 _default_dlopen_flags = sys.getdlopenflags()
     44 sys.setdlopenflags(_default_dlopen_flags | ctypes.RTLD_GLOBAL)
---> 45 from tensorflow.python import pywrap_tensorflow
     46 sys.setdlopenflags(_default_dlopen_flags)
     47 

ImportError: cannot import name pywrap_tensorflow

Is there an additional change needed to my python or ubuntu/bash environment?

Answers:

From the path in your stack trace (/git/tensorflow/tensorflow/…), it looks like your Python path may be loading the tensorflow libraries from the source directory, rather than the version that you have installed. As a result, it is unable to find the (compiled) pywrap_tensorflow library, which is installed in a different directory.

A common solution is to cd out of the /git/tensorflow directory before starting python or ipython.

Questions:
Answers:

I compiled, installed TensorFlow from source (GitHub: https://github.com/tensorflow/tensorflow) in a Python 2.7 virtual environment (venv). It worked fine, but I needed (as others have mentioned, e.g. user “mrry” at Error running basic tensorflow example) to cd out of the partition in which I had compiled TensorFlow, to another partition, in order to be able to import tensorflow into Python. Otherwise, I got various errors, depending on which (source partition) directory I was in. To clarify:

         source: /mnt/Vancouver/apps/tensorflow
can't import tf: Python launched in any of /mnt/...
  can import tf: Python launched in /home/victoria/...

I later just followed the instructions here,

https://github.com/tensorflow/tensorflow/blob/master/tensorflow/g3doc/get_started/os_setup.md#anaconda-installation

and everything just works, fine.

For reference, I am operating on

  • Arch Linux [4.6.3-1-ARCH] x86_64
  • Intel i7-4790
  • xfce 4.12 desktop environment

INSTALL STEPS:

Modify paths, venv names as per your preferences.

  1. Create tf-env:

    cd /home/victoria/anaconda3/envs
    
    conda create -n tf-env python=2.7 anaconda
    

Note: appending the ‘anaconda’ metapackage installs ALL Anaconda packages (NumPy; …).

  1. Source activate that venv (tf-env)

    source activate tf-env
    

Note: added to ~/.bashrc as an alias:

alias tf='echo "  [TensorFlow in Anaconda Python 2.7 venv (source activate tf]" && source activate tf-env'
  1. Install TensorFlow in the tf-env venv:

    (tf-env)$ conda install -c conda-forge tensorflow
    

THIS obviates the use of pip (including the *.whl install script), an alternative approach to installing TensorFlow that is described here:

https://github.com/tensorflow/tensorflow/blob/master/tensorflow/g3doc/get_started/os_setup.md).

WORKS!

(tf-env)[[email protected] ~]$ P

  [P: python]
Python 2.7.12 |Anaconda 4.1.1 (64-bit)| (default, Jul  2 2016, 17:42:40)
[GCC 4.4.7 20120313 (Red Hat 4.4.7-1)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
Anaconda is brought to you by Continuum Analytics.
Please check out: http://continuum.io/thanks and https://anaconda.org

>>> import tensorflow
>>> print tensorflow.__version__
0.9.0
>>> [Ctrl-D]

(tf-env)[[email protected] ~]$

You may then work with TensorFlow in the tf-env; e.g., in Jupyter notebooks launched in this (tf-env) venv.

Questions:
Answers:

I resolve the problem.
Try the follow command:

pip install --upgrade pip

Leave a Reply

Your email address will not be published. Required fields are marked *