Explanation¶

Understanding-oriented discussion of how Yohou-MLflow works and why it is built the way it is. Read these when you want context and background rather than instructions.

• About Saving Forecasters: A yohou forecaster that runs in production is rarely fitted and thrown away. A scheduled job loads the latest saved forecaster, gives it the newest observations, forecasts, and saves it again for the next run, so a single saved forecaster lives for weeks or months. This page discusses what that asks of a saved model, and why Yohou-MLflow answers the way it does. It grew out of yohou#214, which describes that loop.
• Security: Yohou-MLflow is built on a hardened, security-first release pipeline. This page explains the measures that protect you as a user of the package and what each one means in practice, so you can depend on it with confidence.

For what things do rather than why, see Reference.