天天看点

Ubuntu14.04下TensorFlow的安装及配置

在安装了cuda7.5和更新了cudnn版本之后,就可以进行TensorFlow的安装配置了,进入官方下载地址进行下载安装,tensorflow的github地址:https://github.com/tensorflow/tensorflow。

具体安装只需一条指令,

sudo pip install /home/lc/Downloads/tensorflow_gpu-rc0-cp27-none-linux_x86_64.whl
           

更换你自己的路径与文件名

此时安装路径为

/usr/local/lib/python2.7/dist-packages/

这不是本文要介绍的重点,依然要针对我安装过程时出现的问题,进行相关的展开。

Traceback (most recent call last):
  File "<stdin>", line , in <module>
  File "/home/ubuntu/anaconda2/lib/python2.7/site-packages/tensorflow/__init__.py", line , in <module>
    from tensorflow.python import *
  File "/home/ubuntu/anaconda2/lib/python2.7/site-packages/tensorflow/python/__init__.py", line , in <module>
    from tensorflow.python import pywrap_tensorflow
  File "/home/ubuntu/anaconda2/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow.py", line , in <module>
    raise ImportError(msg)
ImportError: Traceback (most recent call last):
  File "/home/ubuntu/anaconda2/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow.py", line , in <module>
    from tensorflow.python.pywrap_tensorflow_internal import *
  File "/home/ubuntu/anaconda2/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow_internal.py", line , in <module>
    _pywrap_tensorflow_internal = swig_import_helper()
  File "/home/ubuntu/anaconda2/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow_internal.py", line , in swig_import_helper
    _mod = imp.load_module('_pywrap_tensorflow_internal', fp, pathname, description)
ImportError: libcusolver.so: cannot open shared object file: No such file or directory
           

当我Python命令行中输入语句

出现以上错误,查了很多别人的疑问解答,都是以改变环境变量为主,

sudo vim ~/.bashrc
export LD_LIBRARY_PATH=”$LD_LIBRARY_PATH:/usr/local/cuda/lib64”
export CUDA_HOME=/usr/local/cuda
source ~/.bashrc
           

但我这么做以后依然没有解决,

仔细分析,其实不光是环境变量路径的问题,TensorFlow现在的安装版本适配的应该是cuda8.0的,而我安装的是cuda7.5版本,所以需要降级版本安装,

可能有些人的电脑pip安装还会出现一些问题,在安装之前最好执行以下语句,

这一条是针对的Python2版本,如果你是Python3(sudo apt-get install python3-pip )

安装之后,运行TensorFlow的运行检测,如果正常就说明万事大吉了。

我的是还遇到了下面的问题,

>>> import tensorflow as tf
Traceback (most recent call last):
  File "<stdin>", line , in <module>
  File "/usr/local/Cellar/python/2.7.6/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/__init__.py", line , in <module>
    from tensorflow.python import *
  File "/usr/local/Cellar/python/2.7.6/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/python/__init__.py", line , in <module>
    from tensorflow.python.platform import test
  File "/usr/local/Cellar/python/2.7.6/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/python/platform/test.py", line , in <module>
    import mock                # pylint: disable=g-import-not-at-top,unused-import
  File "/usr/local/Cellar/python/2.7.6/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/mock/__init__.py", line , in <module>
    import mock.mock as _mock
  File "/usr/local/Cellar/python/2.7.6/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/mock/mock.py", line , in <module>
    from pbr.version import VersionInfo
ImportError: No module named pbr.version
           

针对这个问题,你只需要

sudo pip install pbr funcsigs
           

即可修复。

以上皆是我在安装时遇到的各种问题,通过搜索相关资料,最后都进行了有效的解决,希望能给你有用的参考。

继续阅读