Skip to content

Latest commit

 

History

History
117 lines (67 loc) · 4.73 KB

install-docker-local.md

File metadata and controls

117 lines (67 loc) · 4.73 KB
Raw

How-to Install TensorFlow using Docker on your laptop

These instructions will guide you through downloading and running the TensorFlow Docker container. They'll also show you how to start a Jupyter notebook server, and connect to it from your browser. There's a lot of text ahead, but only a few commands to actually run.

We will present the tutorial in Jupyter notebooks. To run them on your laptop machine, you will need a working TensorFlow installation (v1.3+).

Step 1: Install Docker on your laptop

Visit www.docker.com to download and install Docker on your laptop. Docker is available for Linux, Mac, and Windows. To learn more about Docker in general, you can see the getting started guide - or you can proceed fearlessly ahead.

Step 2: Download the container image

Once Docker is installed and running on your laptop: open a terminal, and download the TensorFlow container using this command.

$ docker pull tensorflow/tensorflow:1.3.0-devel

Note: this is about 800 MB. If you have git installed on your local machine, you can use "tensorflow/tensorflow:1.3.0" rather than the devel build, for a smaller download.

You can find a list of tensorflow containers on Dockerhub here.

Step 3: Create a shared directory

During the workshop, you may wish to copy data between your laptop and the running container. By default, the container's filesystem is isolated from your machine. Here, we will create a directory on your laptop, and mount it as a volume when you start the container. Any data inside this folder will be accessible to both your laptop and the container.

First, let's create the directory we'll mount in your home folder. Feel free to use a different path.

$ mkdir ~/shared

Step 4. Start the container

Run this monster command to start the container.

$ docker run -it -p 8888:8888 -p 6006:6006 -v ~/shared:/shared tensorflow/tensorflow:1.3.0-devel bash

Long command, right? Here's what the arguments do:

  • -v ~/shared:/shared mounts ~/shared on your laptop as /shared inside the container.

  • -p 8888:8888 forwards port 8888 on your laptop to port 8888 on the container. We'll use this to connect to the Jupyter Notebook server.

  • -p 6006:6006 forwards port 6006. We'll use this to connect to TensorBoard.

  • -it attaches your terminal to the container.

  • bash starts bash.

At this point, the container will start - and your terminal prompt will change to resemble something like this.

root@83bf8a86752e:/notebooks#

This means your terminal is connected to the running container. Any commands you run in this window will be run inside the container. Try ls for example, and you will see the containers filesystem.

If you'd like to exit the container and return to your usual terminal, type exit. (You can start the container again with the above command).

Step 5. Clone the workshop

All further commands will be run from inside the terminal connected to the running container.

We will now clone today's workshop inside your shared directory.

# cd /shared
# git clone https://github.com/random-forests/tensorflow-workshop.git
# cd tensorflow-workshop

Step 6. PIP install a few libraries

We will now use pip to install a few libraries. Because we're running inside the docker container, these will not affect your system at all.

# pip install -r setup/requirements.txt

Step 7. Start a Jupyter Notebook server

We will now start a Jupyter Notebook server running inside the container. We will then connect to it using a web browser on your laptop.

Step 7a) Start the notebook server.

From inside the container:

# jupyter notebook --allow-root

You will see output on your terminal to indicate the server is running. Later, if you want to stop the notebook server, press Control-C.

Step 7b) Copy the login token

Notice the second from the last line contains a login token. Copy this; you will need it in a moment to connect to the server.

Step 7c) Use a web browser to connect to the server

Open a web browser on your laptop. Enter localhost:8888 in the address bar to connect to the server. Paste the login token you coped in the previous step when prompted.

Open the examples folder, and click on 00_test_install.ipynb. You should be able to run the notebook without issue.

When you're done

Stop the Jupyter Server by pressing Ctrl-C then y.

To exit Docker, type exit from the terminal.

Note: by default, any data you create inside the Docker container will be lost when it exits. To save data, copy it to the shared directory before stopping the container.