Skip to content

Create a model run

A Model Run represents executions of a Model with particular configurations and specific outputs. In general, it is best to start a new Model Run whenever significant changes have been made to the code version or assumptions to preserve history within the graph. However, Model Runs are flexible and users can decide how they want to organize their work. For example, users can create Model Runs associated with each handoff dataset or a single Model Run for all handoff datasets.

There are three main aspects that are specified when initializing a Model Run:

  1. Model code and configurations
  2. Handoff datasets
  3. Handoff tasks

PIPES provides a utility method for generating a template for the Model Run data based on the Project Run and Model topology:

$ pipes modelrun template -t modelrun-creation -p your_project_name -r project_run_name -m model_name -o model-run-template.toml

This will output a file called model-run-template.toml which you can edit with specifics about the Model Run. The template is pre-filled in by PIPES, which uses the Level 1 metadata of the handoff’s source Model and downstream Model(s) to identify differences between expected requirements and scenarios, and then requires a transformation for each difference. For example, a difference in the geographic resolution between Model A and Model B for a given handoff will require a geographic resolution transformation. The Model Run Config outlines all expected transformations for each handoff and PIPES uses these details to track progress in Level 2 to ensure that all expected handoff transformations have been completed. See Validation for more explanation on how requirements differences are determined and examples.

Note

Although all handoffs and transformations are included in the template, users can remove any that are not relevant to the work being performed for that Model Run.

In addition to transformations, QAQC and visualization Tasks can also be added to account for expected validation steps made during this step in the pipeline. Completed work will be validated against these in future parts of the workflow. See Progress Tracking to learn more.

For more specifics on the metadata keys and their types in the Model Run template, check out the Model Run Config.

To submit your Model Run configuration to start work you can run the following command:

$ pipes modelrun create -p your_project_name -r project_run_name -m model_name -f your_model_run_template.toml

When you are finished with work on a Model Run, or if you simply just want to end the Model Run work early, you can close it. See Closing a Model Run for more information.