Install TensorFlow easily with our 1-Click Installer

We're happy to announce that we've created a 1-Click TensorFlow installer. It allows you to spin up a new server, which is all set up to start coding using TensorFlow.

Learning Tensorflow Is Much More Fun than Installing It

I've talked to so many people who are trying to learn machine learning, and TensorFlow is the natural option that most people want to use. TensorFlow is great, and a great choice for beginners to ML, but I know from experience (mine, and several people I've helped) that getting the tooling all set up can be a real challenge. The Chunkhost 1-Click TensorFlow image allows you to just create a virtual private server with everything installed, and get started on your code.

How to Create a TensorFlow Server

Step 1: Create a new server

Go to your "Create a New Chunk" page, and pick a name for your chunk (if you're new to Chunkhost, we call servers "chunks").

Make sure to choose "TensorFlow" as your OS Image

Click the button that says "Use a Different Image" so that you can choose our 1-Click Tensorflow image, and then you can choose any other options you want (having a sufficient amount of RAM and VCPU is very beneficial for any sort of machine learning models)

That's it

Now, just wait for the server to be set up, and in a couple of minutes, you'll receive an email with the details (the IP address, and username/password if you're not using SSH).

How to Use your TensorFlow Chunk

You'll be using Jupyter Notebook to program with TensorFlow on your server (this is the most popular way to use TF.) Jupyter Notebook is a very simple editor that you run in a browser, so you don't need to install anything on your computer (especially handy if you're planning to use TensorFlow with a Chromebook.)

Jupyter runs on your Chunkhost server, and you connect with any browser, but there is one step you must do first: Setting up an SSH tunnel. This is a way for programs running on a given port on your remote chunk to appear on a specific port on your computer (sorry, bad description, but it's hard to sum up, and not that important for this tutorial.)

Setting up an SSH Tunnel

To set up an SSH tunnel, you will need to know your username, password and IP address. These are all provided in the initial email you receive. The method you'll use depends on your OS.

Step 1a (Mac/Linux): Setting up your SSH Tunnel on macOS or Linux

If you're using Mac or Linux, you're in luck, because there's just one command to create your tunnel. On Mac, open up the app named Terminal, and on Linux open a terminal window. Now type this, but replace username with your actual username, and change the 123.456.78.90 to your chunk's IP:

ssh -L 8000:localhost:8888 username@123.456.78.90

Step 1b (Windows): Setting up your SSH Tunnel on Windows

On Windows, you need a small program called PuTTY, which you can get here. Once you have downloaded and installed it (a super fast process), open it up.

First, you need to enter your IP address in the Host Name (or IP Address) area, as seen in the image below:

Next, click the SSH button on the left, then you'll see an option for Tunnels you should click. Then enter Source port: 8000, Destination: localhost:8888 and then click Add. This is better explained in this diagram:

Now, that you've added all the necessary settings, we're going to save this configuration, so you never have to do this again. Click the word Session at the top and then typed something like Tensorflow Tunnel Setup in the Saved Sessions box, and click Save, like so:

Now, whenever you want to connect to your server (chunk) using this SSH tunnel, you just double-click on the name in the Saved Sessions box. Try that now and go to Step 2:

Ready to code!

Whether you use Mac, Windows, or Linux, you should now see a login window to your server. Enter your username and password in there when it asks you.

Once you do that, you'll get a command prompt that will look something like this:

root@server-name:~#

Type:

jupyter notebook --allow-root

Now, the server will send back a few lines of information, including a token. Copy this token, and open this address in your browser:

http://localhost:8000/

It will ask for your token, so enter it there, and you should be ready to go (for more instructions on running Jupyter and TensorFlow, check out the Tensorflow website, there are some good tutorials there). Leave the server window open until you are done using Jupyter, because once you close it, your SSH tunnel will also terminate.

Enjoy using your TensorFlow server, and let us know if you have any trouble, or have suggestions or questions!

Software Included in This Image

Package	Version
TensorFlow	2.0.0
TensorBoard	2.0.0
Python 3	3.6.7
Jupyter Core	4.6.0
Jupyer Notebook	6.0.1
Pip	19.3
Numpy	1.17.2
SciPy	1.3.1
Matplotlib	3.1.1
Pandas	0.25.1
OpenCV-Python	4.1.1.26
Virtualenv	16.7.6
Keras-Applications	1.0.8
SymPy	1.4
Screen	4.6.2
Curl	7.58.0
Git	2.17.1