GitHub

Running Models

This guide covers how to run models, configure solvers, and work with scenarios. For structural modeling steps, see the Modeling guide.

The JCGEExamples package provides reference models ported from the literature. Each model exposes a model, baseline, and solve entry point.

using JCGEExamples

result = JCGEExamples.StandardCGE.solve()

Build a spec and run it

You can also run a custom RunSpec directly once it is validated:

using JCGERuntime, Ipopt

report = validate_model(spec)
report.ok || error("Model validation failed")

result = run!(spec; optimizer=Ipopt.Optimizer)

Solvers

  • NLP models: Ipopt (Ipopt.Optimizer)
  • MCP models: PATHSolver (PATHSolver.Optimizer)

PATHSolver requires a license for larger problems. You can provide the license via environment variable:

export PATH_LICENSE_STRING="<LICENSE STRING>"

If the problem is small enough, PATHSolver can run without a license.

Runtime options

Typical runtime controls include:

  • Choice of optimizer and solver parameters.
  • Output directories and report formats.
  • Scenario labels for comparison workflows.

Exact options depend on the runtime package entry points you use.

Scenarios and experiments

Run the same model under multiple parameter sets to compare outcomes. A common workflow is to:

  1. Define a baseline calibration.
  2. Apply policy or shock parameters.
  3. Solve and compare outputs.

Use consistent naming for scenarios to make reports reproducible.

Troubleshooting

  • Check validation reports before solving.
  • Confirm that all calibration outputs map to block parameters.
  • For PATHSolver issues, confirm the license string and solver availability.

Next steps