Skip to content

Latest commit

 

History

History
105 lines (74 loc) · 2.94 KB

install-mac-native.md

File metadata and controls

105 lines (74 loc) · 2.94 KB
Raw

Installing TensorFlow Natively on OS X

We will present the tutorial in Jupyter notebooks. To run them on your machine, you will need a working TensorFlow installation (v0.10).

Follow these instructions, which assume you have OS X (probably 10.11 "El Capitan"), and will use Python 2.7.

Open a terminal.

Open Terminal. This tutorial assumes you are using bash, which you probably are.

Clone this repository

Using git, clone this tutorial and enter that directory.

git clone https://github.com/wolffg/tf-tutorial.git
cd tf-tutorial

Install Pip and Virtualenv

Pip is a package management system used to install and manage software packages written in Python. Virtualenv allows you to manage multiple package installations.

At your Terminal window, run the following command.

# Mac OS X
sudo easy_install --upgrade pip

Once you've installed pip, you'll need to add a few more packages.

sudo easy_install --upgrade six
sudo pip install --upgrade virtualenv

These should some dependencies and Virtualenv.

Now, create a virtual environment.

virtualenv --system-site-packages ~/tensorflow

Note: If you have already installed anaconda, some versions of anaconda and virtualenv are not compatible. If you have trouble, such as seeing errors about "sys.prefix", you may want to try to use the TensorFlow anaconda installation instructions. You'll also need to install matplotlib and Pillow as well to get the full experience.

You will need to Activate the environment, which is to say switch your Python enviroment to a fresh one with clean dependencies.

source ~/tensorflow/bin/activate

You are now running in a special Python enviroment with safe dependencies. Your prompt should start with (tensorflow) $.

Select a binary and install protobufs, TensorFlow and Jupyter. For this lab, we will use CPU-only Mac.

# Within the (tensorflow) virtualenv
pip install --upgrade protobuf
export TF_BINARY_URL=https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-0.10.0-py2-none-any.whl
pip install --upgrade $TF_BINARY_URL
pip install --upgrade jupyter
pip install --upgrade Pillow

Running Jupyter

From your "tensorflow" virtualenv prompt, run the following:

(tensorflow) $ jupyter notebook

Click on 0_tf_hello_world.ipynb to test that jupyter is running correctly.

You should be able to run the code in notebook.

Installation notes

Virtualenv is a tidy way of managing your dependencies. Any time you want to run TensorFlow, you can activate by source ~/tensorflow/bin/activate. To exit the virtual environment, simply type deactivate.

Without using Virtualenv, at this time you may run into issues with upgrading some pre-installed Python dependencies (especially numpy on MacOS El Capitan 10.11). Anaconda is another solution to managing dependencies, but the