Tutorials

Tutorials are paced learning sessions. They assume you have already read Quickstart so that installation and stage names are familiar. Start with the short bake-off path, then read the report, then open the detailed notebooks when you need stage-level logs and debugging patterns.

Complete Tutorials after Quickstart. Use Guides when you later have a specific job such as bringing your own traces or building a custom classifier. Use the API Reference when you need exact names.

Getting started

Detailed notebooks

1. First zoo runInstall, link data, launch the public bake-off
2. Read the reportTurn the table into a decision
3. Stage notebooksData, classify, hardware, report
4. DebugLogs, isolation, common failures

Getting started

First zoo run installs the extras used by the public multi-method benchmark, links the paper data the bundled configs expect, and launches configs/paper_all_methods_benchmark.yaml from the command line or from Python.
Read the report opens the summary PDF or HTML and the companion JSON, then explains joint accuracy, F5Q, PQ min, analytical hardware columns, and the recommendation box so you can defend a ranking in a lab meeting.

Detailed notebooks

These Jupyter notebooks live under doc/tutorials/ and are rendered in the HTML manual without re-executing during the docs build. Run them locally from the repository root when you want live logs.

Notebook	Focus
Notebook: data stage in depth	Stage 1 data load, demodulation, splits, caches, and data-stage log lines
Notebook: features and classify stage in depth	Feature kinds, Stage 2 classify banners, isolating one method
Notebook: hardware estimation and report stage	Analytical hardware estimation and summary report outputs
Notebook: debugging the ARCADE pipeline	Log levels, validation, minimal reproductions, symptom → fix map

If a tutorial step fails because an optional dependency is missing, return to Installation and install the matching extra before retrying.