Sagemill: Run SageMaker jobs like papermill

Sagemill lets you to execute your notebook as SageMaker jobs in the similar way to papermill

Overall features

• parameterize notebooks
• execute notebooks as SageMaker Trainig/Processing Job

Installatioin

$ pip install sagemill

Prerequsites

• Python>=3.5

Required IAM policy

• ecr
  • create-repository
  • get-authorization-token
• sts
  • get-caller-identity
• permission to run Sagemaker jobs
  • write to cloudwatch logs
  • access to some S3 buckets
  • start training/proceessing jobs
  • etc.

Example notebooks

These notebook are assumed to be conda_python3 in SageMaker Notebook instance
If you run it on different environments, install conda and run pip install sagemaker
(You cannot see tags in Github. Download and run the notebooks to see tags)

• Training job with tensorflow
• Training job with your own docker image
• Processing job with SKLearnProcessor
• Processing job with your own docker image