dockyard_rl.experience.cua.grading.dispatch

Module Contents

Functions

evaluate_custom

Grade a finished custom episode; return reward in [0, 1].

Data

logger

API

dockyard_rl.experience.cua.grading.dispatch.logger

‘getLogger(…)’

dockyard_rl.experience.cua.grading.dispatch.evaluate_custom(task_config: dict[str, Any], handle: Any, *, judge: Optional[Callable[..., float]] = None) → float

Grade a finished custom episode; return reward in [0, 1].

Args: task_config: the task’s config dict; its evaluator block drives grading. handle: the backend’s live episode handle (duck-typed — evaluators read action_history / eye / cc / host / task_config off it). judge: optional screenshot-judge callable threaded to evaluators that use it.