Udacity EC2 Terraform for Deep Learning
This is a HCL (HashiCorp Configuration Language) project that allows you to create a GPU p2.xlarge instance on AWS for the purpose of training Deep Learning networks and running jupyter handbooks. This will save you creating EC2 instance, security groups etc. manually.
## Usage
in terraform.tfvars replace your aws_profile and aws_region accordingly.
On Unix/Mac OS X you should have a ~/.aws/credentials file that looks like the credentials~sample file
included in this project. If you don't, go to your AWS account and obtain your security credentials, then store
them in the same format as the credentials~sample file.
If you have AWS experience, then you know you can also configure your profile with the aws-cli or with ENV vars.
Replace the profile in terraform.tfvars with the profile you intend to use if you have multiple aws accounts.
### Install terraform
Install terraform here
Make sure to add it to your $PATH.
Run the project
To make sure things are going to plan run terraform plan. If you get no errors go ahead and run terraform apply.
Beware the costs!
If you get errors, and want to avoid runnning a bill, run terraform destroy. Terraform is fantastic software and is very reliable,
but it does not rollback in the face of errors, so make sure to log onto your AWS account and make sure all unused resource are
terminated or destroyed.
### Issues
Sometimes the provisioning (cloning of git repo, cd'ing into folder, starting conda and running jupyter) may not go according to plan. All you need to do is ssh into the machine at the IP address and issue the instructions that failed again. Also, this is only the basic setup, follow the instructions of the particular project/assignment you are working on to download further repos and assets.