This terraform script creates a p2.xlarge instance in AWS and allows you to connect to a Jupyter notebook running on the server. I've tried to make it generic but I've only tested on our own AWS setup so may need some tweaking. Improvements welcome :)

You can choose either the fast.ai AMI or the AWS Ubuntu deep learning AMI

You may need to contact Amazon to get authorisation to run p2 instances, see the fast.ai setup instructions for more details.

• terraform
• autossh

These can be installed with brew on Mac

• terraform init <./fast_ai and/or ./aws_deeplearning>

1. Define variables in terraform.tfvars

• bastion_host: host / ip of bastion host to jump through
• allowed_cidr: cidr block of your vpc, e.g. 10.0.0.0/16

2. Optionally run terraform plan <./fast_ai or ./aws_deeplearning> to check the infrastructure that will be created. Run terraform apply <./fast_ai or ./aws_deeplearning>

3. Run ./connect_notebook.sh

• The Notebook process can take quite a while to start, you may need to wait a minute
• Default password is dl_course

1. Define variables in terraform.tfvars

• allowed_cidr: your public ip address, e.g. "xxx.xxx.x.xx/32"

2. Optionally run terraform plan <./fast_ai or ./aws_deeplearning> to check the infrastructure that will be created. Run terraform apply <./fast_ai or ./aws_deeplearning>

3. Directly connect to <instance_public_ip>:8888

• The Notebook process can take quite a while to start, you may need to wait a minute
• Default password is dl_course

The p2.xlarge currently costs ~$1 per hour.

1. terraform destroy <./fast_ai or ./aws_deeplearning>

• Answer yes / use --force flag