Ansible EC2 data science enviroment

This playbook spins up Amazon EC2 stop instance with preinstalled packages useful for data science and machine learning related tasks:

• miniconda
• zsh
• vim
• tmux
• git
• CUDA and CUDNN (optional)

Provisioning spot instances from scratch

1. Setup ansible for use with Amazon EC2: https://aws.amazon.com/ru/blogs/apn/getting-started-with-ansible-and-dynamic-amazon-ec2-inventory-management/
2. Edit EC2 variables in ec2.yml playbook for your preperences (instance type, spot price, ...)
3. Spin it up with provision.sh

Configuring provisioned instances

Use ec2-gpu-cuda.yml playbook. By default it will apply to all g2.2xlarge instances (see hosts in the playbook).

References

This playbook uses the following roles:

1. https://github.com/robinandeer/ansible-miniconda (MIT License)