Skip to content

Checkpoint: Pipeline Design

Having completed the lessons in this module, you should now be able to do the following comfortably:

  1. Identify code in a notebook that implements a discrete step in a machine learning workflow and annotate that cell as a Pipeline Step.
  2. Identify the data that one step produces as output and the step or steps that depend on that data as input.
  3. Specify single-step and multi-step dependency relationships between Kubeflow pipeline steps using the Depends on parameter of the Pipeline Step annotation.
  4. Create pipeline branches that can run in parallel using the Depends on parameter of the Pipeline Step annotation.
  5. Organize the Python statements that import modules your pipeline steps need into a small number of cells and mark those cells using the Imports annotation.
  6. Identify cells in a notebook that should be excluded from pipeline runs and annotate them as Skip Cells.
  7. Organize and annotate Functions cells.