在centos6环境下安装TensorFlow

The official-released binary packages of TensorFlow are built for newer version of Linux distros. Here is how to build TensorFlow binary package for CentOS 6.

Install Dependencies

Admin privilege is required here.

OpenJDK (for Bazel)

Install OpenJDK

Shell

1	yum install java - 1.8.0 - openjdk - devel

Developer Toolset (for Bazel and TensorFlow)

C++11 compatible compiler is required during the building process. Follow the instructions in this article to install Developer Toolset. It is worth noting that, in this case, installing just devtoolset-6-toolchain would be sufficient.

Python 2.7 (for TensorFlow)

This step also requires Software Collections installed. Therefore, if you’ve installed Developer Toolset by following the instructions above, you should have no problem installing these dependencies.

Install Python 2.7

Shell

1	yum install python27 python27 - numpy python27 - python - devel python27 - python - wheel

Build and Install Bazel

Enter Build Environment

Enter Software Collection environment with Developer Toolset.

Setup Build Environment

Shell

1	scl enable devtoolset - 6 bash

Prepare Source Code

Method 1 (Recommended)

The standard way of compiling a release version of Bazel from source is to use a distribution archive. From version 0.4.1, bootstrapping Bazel from GitHub repository is no longer supported as only the distribution archives contain generated artifacts required for direct compilation. Download bazel-<VERSION>-dist.zip from the release page.

Preparation

Shell

1 2 3 4

# download distribution archive wget https : / / github .com / bazelbuild / bazel / releases / download / 0.4.5 / bazel - 0.4.5 - dist .zip unzip bazel - 0.4.5 - dist .zip - d bazel - 0.4.5 - dist cd bazel - 0.4.5 - dist

Method 2 (Only Works for Version <= 0.4.0)

Preparation

Shell

1 2 3 4 5 6

# clone source code repository git clone https : / / github .com / bazelbuild / bazel .git cd bazel   # select version (optional) git checkout 0.4.0

Build and Install

Build and Install

Shell

1 2 3 4 5 6 7 8 9

# compile . / compile .sh   # install mkdir - p ~ / bin cp output / bazel ~ / bin /   # exit from Software Collection environment exit

Build TensorFlow with Bazel

Enter Build Environment

Enter Software Collection environment with Developer Toolset and Python 2.7.

Setup Build Environment

Shell

1	scl enable devtoolset - 6 python27 bash

Prepare Source Code

Preparation

Shell

1 2 3 4 5 6

# clone source code repository git clone https : / / github .com / tensorflow / tensorflow .git cd tensorflow   # select version (optional) git checkout v1 . 1.0

Since GNU C library version in CentOS 6 is less than 2.17, a slight modification needs to be applied before compilation.

Modify tf_extension_linkopts function in tensorflow/tensorflow.bzl from

Original tf_extension_linkopts Function

Python

1 2	def tf_extension_linkopts ( ) :    return [ ]    # No extension link opts

to

Modified tf_extension_linkopts Function

Python

1 2	def tf_extension_linkopts ( ) :    return [ "-lrt" ]

Build

After above modification, we can now start building!

Remember to disable jemalloc when configuring — otherwise the build will fail.

Build

Shell

1 2 3 4 5 6 7 8 9

# configure workspace, disable jemalloc and leave everything else as default . / configure   # build ~ / bin / bazel build -- config = opt / / tensorflow / tools / pip_package : build_pip_package bazel - bin / tensorflow / tools / pip_package / build_pip_package / tmp / tensorflow_pkg   # exit from Software Collection environment exit

Finishing Up

Finally, you will found a tensorflow-1.1.0-cp27-none-linux_x86_64.whl (filename may vary upon versions) in /tmp/tensorflow_pkg/. With this self-built binary package, we can now deploy TensorFlow to CentOS 6 by pip.

Enter Python Environment on Deployment Target

Enter Python 2.7 environment either by Software Collection or by tools like pyenv.

Enter Python 2.7 Environment

Shell

1 2 3 4 5

# by Software Collection scl enable python27 bash   # or by pyenv pyenv shell anaconda2 - 4.3.1

Install

Depending on your environment setup, admin privilege might be needed here.

Pip Installation

Shell

1	pip install tensorflow - 1.1.0 - cp27 - none - linux_x86_64 .whl

References

1. Installing TensorFlow from Sources – TensorFlow
2. Compiling from source – Installing Bazel – Bazel
3. Distribution Artifact for Bazel – Bazel
4. ImportError: undefined symbol: clock_gettime #121