eni-mlops-content

Support package for the ENI MLOps courses (DEV25_0504B, DEV25_0505B). It holds the code the labs' Kubeflow components import, so a component body stays a few lines of orchestration around a function that is versioned, installed and testable on its own.

What it provides

• mlops_content.config — dataset paths, and the Feature, ModellingStrategy, SplitStrategy and ModelAlias enumerations the pipelines are parameterised by.
• mlops_content.pipeline_steps — load, build_features, split_by_period, train, evaluate, select_best_alpha, predict for the electricity-consumption forecasting the courses use.
• mlops_content.exploration — the plots and summaries of the exploratory notebook.
• mlops_content.drift_metrics — PSI between a reference window and a recent one, pooled across clients and per client, via Evidently.

Install

pip install eni-mlops-content            # the modelling and drift code
pip install "eni-mlops-content[components]"   # plus kfp, mlflow and what a component pod needs

The same wheel goes into the course's component image, which is what makes the labs' component bodies runnable outside a pod: the tests import exactly what the pod imports.

This is course material rather than a general-purpose library — the API follows what the labs teach, and it changes with them.