2.0.4

• Added feature selection to the notebook

2.0.3

Updated preprocess_df in the template notebook

2.0.2

• Added the pre processing functionality
• Formatted some terraform and Python code
• Added unit tests for split_data and preprocess_df

2.0.1: Merge pull request #4 from crederauk/fix/module-issues

Updated retraining_schedule validation

[2.0.0] - 21/12/23

[2.0.0] - 21/12/23

BREAKING CHANGES

• Mandatory variable resource_naming_prefix has now been added.
• Mandatory variable data_s3_bucket has now been added.
• Variable model_instance_count has been renamed inference_instance_count
• Variable sagemaker_instance_type has been replaced with two variables:
  • sagemaker_training_notebook_instance_type, which now defines the type of the training notebook only.
  • inference_instance_type, which now defines the type of instance used to serve the model only.
• The syntax for data_location_s3 has now changed - should not include bucket name.
• region and account_id variables have been removed - no longer required.
• All outputs have now been refactored to be more useful.
• When creating a Docker container from the image in this module, MODEL_NAME and MODEL_TYPE must now be passed in as environment variables.
• random provider is now required to generate suffix for globally unique bucket names
• Permissions have been locked down:
  • Lambdas can no longer assume the created Sagemaker IAM role by default.
  • Glue IAM role and Sagemaker IAM role no longer have access to all KMS keys in the account. They are only given access to required S3 bucket encryption keys.
  • iam:PassRole permission has been removed from the Sagemaker IAM Role