Designing a DVC pipeline

Designing a DVC pipeline, or DAG, is fundamental to leveraging DVC in your machine learning workflows. DAGs allow us to codify inputs, outputs, and execution of a certain step. The outputs of one step can serve as input to one or more steps, thereby naturally setting the right dependencies between steps.

In this exercise, you'll work on designing an ML workflow that contains four stages, namely,

• Data preprocessing (preprocess_stage)
• Data splitting (split_stage)
• Model training (train_stage)
• Model evaluation (evaluate_stage)

We will exclusively work with the dvc stage add commands. Scroll down to the end of the shell script file (dvc_dag_stages_add.sh) if needed.