In-depth Configuration

Configure pipelines at will with ZenML.

This is an older version of the ZenML documentation. To read and view the latest version please visit this up-to-date URL.

The goal of this section is to showcase some critical advanced use-cases regarding the configuration of different ZenML resources.

List of topics

• Runtime Settings showcases how to configure different ZenML resource during the runtime of a pipeline.

• Passing Custom Data Types through Steps via Materializers is required if one of your steps outputs a custom class or other data types, for which materialization is not defined by ZenML itself.

• Specifying Hardware Resources for Steps explains how to specify hardware resources like memory or the amount of CPUs and GPUs that a step requires to execute.

• Access metadata within steps via Step Fixtures can, for instance, be used to load the best performing prior model to compare newly trained models against.

• Controlling the Step Execution Order explains how to control the order in which steps of a pipeline get executed.

We will keep adding more use-cases to the advanced guide of ZenML. If there is a particular topic you cannot find here or any use-case that you would like learn more about, you can reach us in our Slack Channel.