maple.backend.policy.base.PolicyBackend.act

abstractmethod PolicyBackend.act(handle: PolicyHandle, payload: Any, instruction: str, model_kwargs: Dict[str, Any] | None = {}) → List[float]

Get action prediction from the policy.

Must be implemented by subclasses to provide inference logic. Takes transformed observations and returns predicted actions.

Parameters:

• handle – Policy handle for the running container.

• payload – Transformed observation from environment (post-adapter).

• instruction – Natural language instruction for the task.

• model_kwargs – Model-specific parameters

Returns:

Predicted action as list of floats.