Model routing
Model routing: Dynamic selection of a model or execution path based on request, cost and constraints. A router can distinguish simple and complex tasks, data locality or availability. Deterministic rules and classifiers must be monitored to prevent savings that degrade quality.
In italiano: Model RoutingLegge IV · Loss, gradiente e ottimizzatoriA router can distinguish simple and complex tasks, data locality or availability. Deterministic rules and classifiers must be monitored to prevent savings that degrade quality.
Examples
- Small model for FAQs
- Fallback after failure
- Routing by data sensitivity