Deep Learning sandbox Cloud Formation Template
deeplearning-env-cfn
Includes 2 variants of DL environment, EFS persistent storage, Spot Instances support. Completely fits under free tier when using t2.micro and less than 5GB of EFS space.
Currently has a choice of two environments:
- ubuntuBootstrapAMI – Bootstraps environment based on Ubuntu 16.04, Anaconda 2-4.2.0, Cuda 8.0.44, Keras 1.2.2 and Theano
Root volume – 16Gb, Data – EFS mounted under /home/ubuntu/efs
Compatible to run fast.ai courses notebooks.
- awsDeepLearningAMI – Amazon Deep Learning AMI 2.2_Jun2017
Root volume – 50Gb, Data – EFS mounted under /home/ec2-user/efs
In case of custom AMIs – EFS is not mounted automatically and Jupyter password parameter is ignored.
Regions to AMIs mapping:
| Image | us-east-1 | us-west-2 | eu-west-1 |
|---|---|---|---|
| awsDeepLearningAMI | ami-e47723f2 | ami-c6dfb2a6 | ami-df0a1ab9 |
| ubuntuBootstrapAMI | ami-80861296 | ami-efd0428f | ami-a8d2d7ce |
Prerequisites
- At least 1 AWS VPC, subnet and security group.
- At least 1 EC2 Key Pair.