CentOS 7系统安装Tensorflow框架

参考教材

知识库参考链接：https://www.mtyun.com/library/45/how-to-install-tensorflow-on-centos7/

官网安装方法（推荐）：https://tensorflow.google.cn/install/install_linux

Tensorflow对CUDNN对版本兼容性

1.3.0：

All our prebuilt binaries have been builtwith cuDNN 6. We anticipate releasingTensorFlow 1.4 with cuDNN 7.

1.2.0:

TensorFlow 1.2 may be the last time we build withcuDNN 5.1. Starting with TensorFlow 1.3, we will try to build all our prebuilt binaries with cuDNN 6.0. While we will try to keep our source code compatible with cuDNN 5.1, it will be best effort.

1.1.0:

TensorFlow 1.1.0 will be the last time we release a binary with Mac GPU support. Going forward, we will stop testing on Mac GPU systems. We continue to welcome patches that maintain Mac GPU support, and we will try to keep the Mac GPU build working.

软件准备

Nvidia 驱动：sh  NVIDIA-Linux-x86_64-375.66.run

Tensorflow：https://tensorflow.google.cn/install/install_linux#the_url_of_the_tensorflow_python_package

cuDNN v5.1 for CUDA8.0

tensorflow_gpu-1.1.0-cp27-none-linux_x86_64.whl

S3下载地址：http://tonydong-49061403.mtmssdn0.com/NVIDIA.375.66.tar & http://tonydong-49061403.mtmssdn0.com/NVIDIA-Linux-x86_64-384.66.run

官方下载地址：

Nvidia驱动下载：http://www.nvidia.cn/Download/index.aspx?lang=cn

CUDA Toolkit Download：https://developer.nvidia.com/cuda-downloads

CUDA安装及兼容性：http://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html

下载tensorflow：https://tensorflow.google.cn/install/install_linux#the_url_of_the_tensorflow_python_package

根据官网给出的链接，修改版本即可，将1.3.0改成1.1.0，就可以下载老版本的。

如https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.3.0-cp27-none-linux_x86_64.whl，

改为https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.1.0-cp27-none-linux_x86_64.whl

查看TF的版本及安装路径：

python

>>> import tensorflow as tf

>>> tf.__version__

>>> tf.__path__

安装过程中报错解答：

GPU driver 与 CUDNN／CUDA不匹配，降低 driver版本或者升级CUDNN/CUDA，在执行python时，import tensorflow as tf 报错

ImportError:libcusolver.so.8.0: cannot open shared object file: No such file or directory

ImportError: /usr/lib/python2.7/site-packages/tensorflow/python/_pywrap_tensorflow_internal.so: undefined symbol: cudnnConvolutionBiasActivationForward

sess = tf.Session()时找不到GPU，没有安装tensorflow-gpu，或安装中有错误，重新安装

>>> sess = tf.Session()

2017-09-13 18:10:26.267041: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.1 instructions, but these are available on your machine and could speed up CPU computations.

2017-09-13 18:10:26.267091: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.2 instructions, but these are available on your machine and could speed up CPU computations.

2017-09-13 18:10:26.267107: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX instructions, but these are available on your machine and could speed up CPU computations.

2017-09-13 18:10:26.267119: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX2 instructions, but these are available on your machine and could speed up CPU computations.

2017-09-13 18:10:26.267132: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use FMA instructions, but these are available on your machine and could speed up CPU computations.

>>> print(sess.run(hello))

TensorFlow 版本太高，CUDNN不支持，需要降低TF或者升级cuDNN

ImportError:libcudnn.so.6: cannot open shared object file: No such file or directory

参考：https://www.jianshu.com/p/10542140e2e3