Mac OSX 上安裝 TensorFlow [CPU support only]

本文介紹在 Mac OSX 系統(tǒng)上如何安裝 Tensorflow ，但除了操作系統(tǒng)包管理有差異，其它內容使用于其它操作系統(tǒng)。

TensorFlow 可以在 Python 2 中運行，但，Python 3 才是未來。所以，建議大家直接使用 Python 3！

注：本文安裝的是 TensorFlow with CPU support only ；電腦上沒有NVIDIA 顯卡，所以我理解應該沒法安裝 TensorFlow with GPU support。

安裝 Python 3

$ brew install python3

安裝 virtualenv

安裝 virtualenv

$ sudo pip install -U virtualenv virtualenvwrapper

將下面命令加入 ~/.bashrc 或 ~/.zshrc，比如我使用zsh，所以加到 ~/.zshrc 文件末尾:

$ echo 'test -f /usr/local/bin/virtualenvwrapper.sh && source /usr/local/bin/virtualenvwrapper.sh' >> ~/.zshrc

安裝 TensorFlow

$ mkvirtualenv -p python3 tensorflow
$ pip install -U tensorflow

安裝 IPython

IPython 是一個體驗特別好的 Python 交互式終端，安裝：

$ workon tensorflow
$ pip install ipython

測試是否安裝成功

$ ipython
In [1]: import tensorflow as tf
In [2]: hello = tf.constant('Hello, TensorFlow!')
In [3]: sess = tf.Session()
In [4]: print(sess.run(hello))
b'Hello, TensorFlow!'

看到成功輸出 b'Hello, TensorFlow!' 說明已經成功安裝 TensorFlow ！

這行之后：

In [3]: sess = tf.Session()

可能會看到類似下面的輸出：

2017-06-14 13:50:49.512831: 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.
2017-06-14 13:50:49.512872: 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.
2017-06-14 13:50:49.512881: 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.
2017-06-14 13:50:49.512894: 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.
2017-06-14 13:50:49.512903: 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.

上面輸出中的 W 表示 警告 (Warning)，提示從源碼編譯并開啟一些編譯選項后可以加快CPU計算速度。

三種辦法避免這類錯誤

上面的 Warning 信息并不影響學習 TensorFlow ，只是會導致 TensorFlow 運行的不夠快。但，如果你還是不希望看到這些 Warning ，可以用下面的三種方法之一。

第一種僅僅是讓你不再看到 Warning，而最后兩種能讓 TensorFlow 運行的更快！

設置 tensorflow log level，避免 warning 輸出

TF_CPP_MIN_LOG_LEVEL

• It defaults to 0, showing all logs
• To filter out INFO set to 1
• WARNINGS, additionally 2
• and to additionally filter out ERROR logs set to 3

$ ipython
In [1]: import os
In [2]: os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
In [3]: import tensorflow as tf
In [4]: hello = tf.constant('Hello, TensorFlow!')
In [5]: sess = tf.Session()
In [6]: print(sess.run(hello))
b'Hello, TensorFlow!'

安裝別人編譯好的 TensorFlow

移步lakshayg/tensorflow-build

更多，請參考：https://github.com/yaroslavvb/tensorflow-community-wheels

自己編譯 TensorFlow

請參考官方文檔：Installing TensorFlow from Sources