Open Models Almanac
Methodsinference & servinginference scheduling

single mention · inference & serving

Runtime-signal-based rollout concurrency auto-throttling

Uses active and queued request counts and KV-cache utilization to dynamically limit requests sent to inference.

The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See inference scheduling for the methods filed beside it.

Also called auto-throttling mechanism.

Filed under inference & serving › inference scheduling