Installation

The steps regarding the environment setup and installation of our scripts will be found here. We will be using Docker for the Tensorflow container. If you want to know more about what Docker is, click here.

Environment setup

As mentioned earlier, we are using an Intel NUC for this project. We started with a clean installation of Ubuntu 18.04.3 LTS and if you don't know how to create bootable usb sticks/install Ubuntu, follow the instructions stated here. Before completing the setup there is an option to install additional software. Select the docker option, because that is where the Tensorflow Docker container will be installed.

If you don't want to install a clean version of Ubuntu, you can just install Docker using the commands below:

• sudo apt-get update && sudo apt-get upgrade installs the latest updates on your system
  
• sudo apt-get install docker.io installs the latest version of Docker
  
• sudo systemctl start docker starts the Docker service in the background.
  
• sudo systemctl enable docker (optional) enables docker to run at the startup of your computer.

After installing Docker you can issue the docker --version command to verify the installed Docker version number.

Container setup

Now we are going to download,install and setup the Tensorflow container. To do this, open a terminal as root and use the command below:

docker run -it --name tensorflowAI
-v $(realpath ~/tensorflow/archive):/tf/archive
-v $(realpath ~/tensorflow/special-project):/tf/special-project
-p 2000:8888 tensorflow/tensorflow:latest-py3-jupyter

The --name option let's you give a name to the container, the -v option lets you add a REAL save location for your files, because you should NOT keep data inside the docker container. The -p option lets you specify a port from which tensorflow will be available outside the container (in this example it's available ouside on port 2000 and the application standard port is 8888). tensorflow/tensorflow:latest-py3-jupyter is the container, which will be used from tensorflow. There are a lot of different Tensorflow container types you could use (click here for more examples), but we will be using the one with python3 and jupyter installed.

After issueing the command connect to your server's ip address (ie. 192.168.1.15:2000) or your local ip address (ie. http://localhost:2000, if you installed it on the same computer) to access the Jupyter environment.

If you look at your console, you will find a token, which will be needed to change the master password to your Jupyter environment. Insert this token into your browser (token field) and enter a new password (You cannot forget this!). After logging in you can see three folders: archive, special-project and tensorflow-tutorials. To check out the tutorials posted by Tensorflow, check the tensorflow-tutorials folder.

The next part of this page will cover the setup of our scripts.

More info coming soon...