Glossary · Infrastructure

Warm Start (Inference)

Serving a request using a model instance that's already loaded into memory and ready, avoiding the latency penalty of a 'cold start' where infrastructure must first initialize.