Skip to content

How-to Guides

Find the operation you need once you have a problem or a result. For a guided learning sequence, start with Tutorials; for exact parameters and guarantees, use Reference.

Define and configure a run

Task Guide
Express objectives, bounds, and constraints Define your problem and check the evaluator.
Check whether an algorithm supports your encoding Algorithm capability matrix and problem catalog.
Supply an algorithm configuration Configuration recipe and configuration fields.
Run from the command line or validate a configuration file CLI & Config and validation recipe.
Use Numba or evaluate in multiple processes Numba recipe and multiprocessing recipe.
Configure stopping and an external archive Stopping & Archive.

Preserve and recover work

Task Guide
Save or load an individual run Run artifacts.
Verify an artifact or reproduce a supported built-in run Inspect and verify, then exact built-in replay.
Plan a persistent experiment matrix Plan the study budget.
Recover interrupted work or retry failed study tasks Resume and retry.
Diagnose a refused replay, invalid value, or missing dependency Troubleshooting.

Loading reads stored data, verification checks an artifact, and replay executes the recorded run. Study resume and retry operate on task state. Follow the linked guide for the guarantees of the operation you need.

Inspect and analyze results

Task Guide
Match decision rows to objective rows Understanding results.
Select one feasible solution using preferences Selection recipe.
Export a table or combine saved fronts DataFrame export and combined fronts.
Compute hypervolume with an explicit reference point Hypervolume recipe and reference-relative ZDT1 example.
Use statistical or plotting helpers Analysis & Visualization, Experimental.

Optional experimental workflows

These destinations state their own dependencies and support boundaries:

  • Hyperparameter tuning: define the scoring, budget, and validation split before using the tuning CLI.
  • Scaling with Dask: distribute problem evaluation when its cost justifies coordination overhead.
  • VAMOS Studio: explore stored results and use trusted local Python workflows.
  • VAMOS Assist: prepare specifications with the template-first flow or an optional provider.

The Cookbook remains the complete recipe index. For runnable scripts and notebooks, use the Examples gallery.