1. Provision the GPU instance using the cloudformation script

aws cloudformation create-stack --stack-name <stack-name> --template-body file://provisioning/ec2-p2-xlarge.yml

2. Make sure your SSH key and SSH agent forwarding is enabled by adding the following lines to ~/.ssh/config

Host <hostname> <host-ip>
    ForwardAgent yes

3. Run the Ansible playbook

ansible-playbook setup-machine-learning-server.yaml -i hosts --user=ubuntu

4. From a different terminal, open an SSH terminal into the EC2 instance

ssh ubuntu@<ip-address-or-hostname> -NL 1234:localhost:1234

5. SSH into the EC2 instance and start Jupyter notebooks

jupyter notebook --no-browser --port 1234

(Cut and paste the login URL with token into your local browser, and you should connect to the Notebook server.)

6. After working on your notebook a little and saving it, make sure to copy it to your local repo and commit changes:

scp ubuntu@<ec2-host-name>:~/machine-learning-notebooks/<notebook-name>.ipynb ./