DeleteMostStaleStrategy iteratively deletes allocated work from stale slots until there is enough memory to allocate the new workload.

ErrorHandlingStrategy is a function that handles errors returned by the scheduler.

Filter filters a slice based on a predicate function and returns a new slice with matching elements.

FilterMap filters a map based on a predicate function and returns a new map with matching entries.

This strategy attempts to spread work across all runners, effectively a min utilization strategy.

NewScheduler creates a new scheduler with a workload allocator.

NewSlot creates a new slot with the given runnerID and work staleTimeout is a function that determines if a slot is stale errorTimeout is a function that determines if a slot has errored.

NewTimeoutFunc returns a function to check if a runner has been idle for a specified timeout duration.

NewWorkloadAllocator creates a new allocator instance with timeout functions for models and runners.

Reverse reverses the order of elements in a slice.

Values returns a slice of all values in the map.

Scheduler is is the main entrypoint to the scheduler package.

WorkloadAllocator defines an interface for managing the allocation of workloads to runners.

TimeoutFunc defines a function type that determines if a runner has timed out based on the last activity.