Bitfusion Documentation

Bitfusion makes it easy, efficient, and elastic to do Deep Learning with GPUs. On this site, you can find our documentation and guides to help you start working with Bitfusion as quickly as possible.

Get Started    

Search results for "{{ search.query }}"

No results found for "{{search.query}}". 
View All Results

AMI User Guide

Who This Guide Is For

This guide is for users of Bitfusion's pre-built Amazon Machine Images on the Amazon Web Services (AWS) Marketplace. Our AMIs do not contain the full capabilities of the Bitfusion Flex platform, however the AMIs do provide a quick way to get started with all the right software and drivers for various deep learning frameworks, scientific computing, and media processing. Check out the variety of AMIs that Bitfusion provides here.

Additional AMI Resources

  • Bitfusion Public Slack Channel - join us here to discuss an questions you may have with regards to Bitfusion AMIs
  • Detailed AMI Documentation - each AMI comes with detailed documentation on latest software inclusions, versions, and more
    This guide assumes you are new to spinning up machines on AWS and will walk through the process of deploying a Bitfusion AMI.

Step 1: Create an AWS Account

If you don’t have one already, make sure you create an account on AWS. (If you already have an account, skip this step and just log in.)

  1. Go to the AWS Console.
    Navigate to, enter your e-mail address or mobile number, and select “I am a New User.”
  1. Enter Login Credentials and Contact Details.
    Enter your name, e-mail, and password, followed by either your company or personal details.
  1. Enter Payment Information.
    There are no upfront fees -- Amazon will only charge based on the cloud infrastructure services that you consume.
  1. Verify your Identity.
    Amazon will call you and ask you to enter the PIN on the screen to ensure that you are who you say you are.
  1. Choose your Support Plan.
    You can pay extra to have an AWS support representative you can call and other service add-ons. Most folks just choose the Basic plan at first, which is free/included -- you only pay for the machines as you consume them. You can always change this later if you need to.
  1. Congrats, you signed up!
    Now login by clicking “Sign In to the Console” or navigating to

Step 2: Deploy a Bitfusion AMI

  1. Create a Key Pair.
    In order to quickly deploy an instance on AWS, you must have a key pair. If you already have a key pair that you plan to use, you can skip ahead to the Bitfusion marketplace page. Go to the EC2 dashboard, either by clicking this link to go there directly, or by going to the AWS Console and then clicking EC2.

Click on Key Pairs in the left sidebar.

Then click on the blue “Create Key Pair” button.

Give it a name. For this example, I will call mine bitfusion-testing. Then hit “Create”.

This will download a .pem file to your Downloads folder (or wherever your web browser has been set to store downloaded files to).

Make sure you take good care of this file.

If you lose track of this file you will not be able to connect to any instances that you create with this key.

Another way to find Bitfusion AMIs

You can also find the Bitfusion AMIs by going to and searching for “Bitfusion”.

  1. Click the AMI to Deploy.
    In this example, we will deploy Bitfusion Ubuntu 14 TensorFlow by clicking on the top most product in the image below. The process will be similar for all other Bitfusion AMIs.
  1. Choose a Data Center Region and Click Continue.
    On the Product Details page you can learn more about the AMI, the pricing approach, and move forward with deployment. Bitfusion provides a free trial period for each of our AMIs for the Bitfusion software, however, you will still be responsible for any infrastructure costs payable to AWS. (This is outside of our control unfortunately -- we'd love to make the AWS servers free too if we could!)

Data center near you

It is a best practice to pick a data center that is regionally close by to your location. For example, if you are on the US East Coast, it is probably best to use US East (N. Virginia) or US East (Ohio). If you are in the UK, it is probably best to use EU West (Ireland) or EU West (London).

  1. 1-Click Launch.
    In order to do a 1-Click launch, you must already have a key pair that you can use. If you have an angry red “Please select a key pair” message as shown in the image below, then you don’t have an available key pair yet, and need to create one. Please Go Back to Step 2 Item 1, where we walk you through how to Create a Key Pair.

If you have an available Key Pair created already, the screen will look like this with a big yellow button prompting you to Accept the Software Terms & Launch, and no red warning messages - But DON'T press the yellow button just yet!

  1. Choose Instance Type.
    In this example, we will select a t2.micro, which is a pretty small, inexpensive instance.

Default instance size

The Bitfusion Ubuntu 14 Tensorflow AMI default is p2.xlarge as GPU instances are optimal for deep learning applications, but for testing purposes you can choose a much smaller instance, such as the t2.micro that we are doing in this example. To get the full power of deep learning frameworks like Tensorflow consider using a GPU for training, and consider using our Bitfusion Flex platform as the best way to build AI!

Request limit increase for GPU instances

Keep in mind -- if you want to deploy P2, G2, or other larger instance types you may need to request a limit increase. If GPU instances don't deploy for you initially, needing to request a limit increase is often the culprit.

  1. Double Check the Key Pair.
    Scroll down to see which key pair is being used. If it did not default to the key pair you want to use for this instance, change it to the one you’d like.
  1. Accept Software Terms & Launch with 1-Click.
    Scroll back to the top and hit the big yellow button to continue with deployment.
  1. Subscription Successful, Instance Deploying.
    Subscription takes a few moments to complete, then your instance deploys immediately thereafter.
  1. Check out your New Bitfusion AMI Instance!
    To watch as your instance comes online, navigate to the EC2 dashboard and select the “Running Instances” link.

Step 3: SSH into your Bitfusion AMI

SSH (secure shell) connections allow you to get to the command line of your AWS instance and run programs there. These instructions will work for Mac or Linux. If you are using Windows, follow these instructions to install PuTTY.

  1. Make sure your instance is running and checks have passed.
    If you aren’t already there, go to the EC2 dashboard and click “Running Instances.”) Instance State should show a green circle and say “running,” and Status Checks should show a green check-mark and “2/2 checks passed.”
  1. Open your favorite Terminal application.
    Navigate to the directory that you saved your .pem file to. For me, it downloaded to my Downloads folder. You may want to stick your .pem files somewhere else, but for this example we’ll just navigate to the Downloads folder. (New to the UNIX command line or a little rusty? There are tons of guides like this one for the basics.)
  1. Make sure your .pem file is not publicly viewable.
    If your .pem file is publicly viewable, SSH will not connect. To change the permissions to make it private, type:
chmod 400 bitfusion-testing.pem

Don't Forget

Don’t forget to replace bitfusion-testing.pem with the name of your .pem file.

  1. Copy your public DNS.
    In the EC2 Dashboard, copy the public DNS address of your instance. You can find it in the Description panel below or by clicking the Connect button.
  1. Run SSH command.
    Go back to your Terminal window, and type:
ssh -i bitfusion-testing.pem

Don't Forget

Don’t forget to replace bitfusion-testing.pem with the name of your .pem file and with Your Public DNS that you copied in the previous step.

The first time you connect, it will ask you "Are you sure you want to continue connecting (yes/no)?" Say yes.

  1. BOOM! Welcome to your Bitfusion AMI.
    Make sure to register with when it prompts you so we can stay in touch and help support you. Check out our blog for additional tutorials coming soon where we’ll use our AMIs to walk through and explore various deep learning and data processing frameworks and use cases.

Additional AMI Resources

  • Bitfusion Public Slack Channel - join us here to discuss an questions you may have with regards to Bitfusion AMIs
  • Detailed AMI Documentation - each AMI comes with detailed documentation on latest software inclusions, versions, and more
    This guide assumes you are new to spinning up machines on AWS and will walk through the process of deploying a Bitfusion AMI.

Step 4: Getting Started with Jupyter (optional)

Jupyter is included in many of our Bitfusion AMIs, and is a great, user-friendly interface for performing many data science functions.

  1. Copy your Instance ID and your Public DNS.
    You can get the Instance ID and Public DNS from the EC2 dashboard by going to “Running Instances,” then selecting your Bitfusion AMI instance and grabbing the info from the Description panel at the bottom of the screen.
  1. Double-check that port 8888 is open.
    Click Security Groups in the left-hand side bar of the EC2 dashboard.

Select the Security Group related to your Bitfusion AMI instance. Then select the “Inbound” tab.

If there isn’t a rule for port 8888, create one now. Click “Edit” then “Add Rule.”

Choose Custom TCP Rule, enter port number 8888, and either choose “My IP” to do your current IP address, or choose “Anywhere” to put and open it up to all IP addresses. (Just be aware that you’ll probably want to lock this down later to be more secure.) Then hit the blue “Save” button.

  1. Navigate to the Jupyter interface.
    Open a tab in the web browser of your choice and navigate to the following URL:

http://{EC2 Instance Public IP}:8888

Except replace {EC2 Instance Public IP} with your Public DNS that you copied earlier.

  1. Enter the Jupyter password.
    Your password is your Instance ID that you copied earlier.
  1. Welcome to the Jupyter interface!
    For most of our AMIs we include some example folders that contain code snippets and explanations of how to run analysis using Jupyter. The Tensorflow AMI contains Udacity course material. You can click the udacity folder to see some example notebooks.

Let’s click 1_notmnist.ipynb.

This is the notebook view. If you’ve used Google Docs in the web browser before, it has a very similar look and feel. Text, code, and images can be weaved together for storytelling of textual explanations, links, code, data visualizations, and more in a single interactive page.

  1. Upload a Notebook.
    Have an existing notebook ready on your personal computer that you’d like to bring into your Bitfusion AMI? Hit the “Upload” button and perform easy file transfers.
  1. Create a New Notebook.
    You can create a new notebook by hitting the “New” dropdown button and selecting either Python 2 or Python 3. For AMIs other than Tensorflow you can start other kernels such at iTorch, Julia, R, etc.

In this example, we’ll create a Python 2 notebook. After that, an Untitled notebook will show up.

We’ll go ahead and name it by clicking the word “Untitled” near the Jupyter logo in the top left, type “Jupyter Testing” in the prompt, and hit “OK.”

Then we can type in some code. For simplicity sake, we’ll just type in:

x = 1 + 1
print x

Hit the Run button to run that green selected code block, which looks like a play/pause icon.

1 + 1 = 2 … it worked!

  1. Start a web shell.
    Want to pull in your own libraries, git-clone some of your code, or perform other command line actions? There’s no need to SSH into the machine, you can just use the Jupyter web shell instead.

From the file browser view, click the “New” dropdown and then select “Terminal”.

  1. Shutdown notebooks or terminal windows.
    Notebooks and terminals are a living, running processes. If later you want to “shutdown” a notebook or terminal, from the file browser view, you can either select the notebook you want to shutdown, then hit the yellow “Shutdown” button, or you can go to the “Running” tab and hit the yellow “shutdown” button to the right of the process you would like to end.

Additional Jupyter Resources

AMI User Guide