class forml.provider.runner.pyfunc.Runner(instance: asset.Instance | None =
None, feed: io.Feed | None =
None, sink: io.Sink | None =
Non-distributed low-latency runner turning the task graph into a single synchronous python function.
This runner is designed for repeated calls for low-volume predictions - the typical use case in online serving.
Upon initialization, the runner instantiates all actors and preloads their states (if stateful) to avoid unnecessary delays when (repeatedly) performing the actual execution.
This runner is internally used by the serving engine. It does not support training/tuning actions. Defining it explicitly using the platform configuration for other runtime mechanisms is not usual.