Machine Learning

Testing Basics

Simple tests that prevent silent ML failures.

What to test first

  • Feature functions: shapes, dtypes, missing value handling.
  • Training smoke test: runs end-to-end on a tiny sample.
  • Inference contract: input schema and output type/range.

Quick checklist

  • Model loads from disk without errors.
  • Predict works with a single row and a batch.
  • Unknown categories / missing columns handled predictably.