1. From the EC2 Dashboard, open the Launch instance dialog.
  2. Label the instance as wished.
  3. Select an OS, for example Ubuntu.
  4. Select an AMI with nVidia drivers contained. Here the Deep Learning Base image is selected as it contains nVidia drivers and toolkit:
    https://docs.aws.amazon.com/dlami/latest/devguide/gpu.html
  5. Select an Instance type. Only certain AWS instance types have GPU. A recommended list can be found on the website:
    https://docs.aws.amazon.com/dlami/latest/devguide/gpu.html
    For this example, we will use a g4dn instance type:
  6. Create your Key pair to access the instance.
  7. Download the .pem file in your browser and save it as you will need it to connect to the instance.
  8. Create your network. You may want to switch SSH traffic from Anywhere to your specific IP. Extra inbound rules can be added to the security group later if needed.
  9. Configure your storage. Be aware that most image models are at least 5GB in size.

    By adding the nVidia OSS image there is an extra volume.
  10. Launch the instance.
  11. When started, use the key pair .pem and the default user ssh inside the instance. Use it's public IP address: ssh -i arivisEC2.pem ubuntu@<ip address>