Scenarios by Domain

Choose the subject of the calculation. Each domain page links the typed scenario contracts and identifies their implementation files.

For a first calculation, start with Choosing a Python API. For a known class name, use Scenario Classes A–Z.

Scenario definitions describe inputs and calculations. Named configurations in scenario_catalog/ select those inputs; result types in result_types/ describe returned values. These packages use the same domain names. The domains follow modeling subjects across all textbook volumes.

• Shared contracts: Common scenario metadata and reusable input conventions.
• Training: Training work, state, updates, and parallel execution.
• Serving: Inference work, request scheduling, and token-cache capacity.
• Communication: Network topology, transfer costs, and collective communication.
• Data: Data preparation, selection, storage, and delivery.
• Reliability: Failure exposure, recovery, and operational safeguards.
• Sustainability: Power, heat, energy, carbon, and material accounting.
• Economics: Cost, resource pricing, staffing, and investment tradeoffs.
• Optimization: Execution bottlenecks, hardware utilization, and resource placement.
• Agents: Model calls, context, tools, and coordinated agent execution.
• Robotics: Physical-machine scenarios, grouped by sensing, control, motion, and learning.
• Statistics: Sampling, uncertainty, evidence, and evaluation decisions.
• Controlled Studies: Typed inputs for bounded, authored teaching studies.