| """Fixed-action harness diagnostic; never a participant model or scored baseline.""" |
|
|
| from __future__ import annotations |
|
|
| from typing import Mapping, Sequence |
|
|
| from benchkit.contracts import ( |
| ActionChunk, ActionSpec, EpisodeContext, JsonValue, Observation, ObservationSpec, |
| PolicyDecision, PolicyDescriptor, |
| ) |
| from benchkit.registry import register_policy |
|
|
|
|
| class ConstantPolicy: |
| def __init__(self, action: Sequence[float], chunk_length: int = 5) -> None: |
| if type(chunk_length) is not int or not 1 <= chunk_length <= 1024: |
| raise ValueError("chunk_length must be an integer in [1, 1024]") |
| self.chunk = ActionChunk.from_steps([action] * chunk_length) |
| self.descriptor = PolicyDescriptor( |
| name="constant-action-diagnostic", kind="scripted", version="1", |
| parameters={"action": list(self.chunk.steps[0]), "chunk_length": chunk_length, |
| "purpose": "infrastructure_validation_only"}, |
| ) |
|
|
| def compatibility_errors(self, observation_spec: ObservationSpec, action_spec: ActionSpec) -> Sequence[str]: |
| try: |
| action_spec.validate(self.chunk.steps[0]) |
| except (ValueError, TypeError) as exc: |
| return [str(exc)] |
| return [] |
|
|
| def start_episode(self, context: EpisodeContext) -> None: |
| pass |
|
|
| def act(self, observation: Observation) -> PolicyDecision: |
| return PolicyDecision(action=self.chunk) |
|
|
| def end_episode(self) -> None: |
| pass |
|
|
|
|
| @register_policy("constant") |
| def make_constant(parameters: Mapping[str, JsonValue]) -> ConstantPolicy: |
| if set(parameters) - {"action", "chunk_length"}: |
| raise ValueError("constant supports only action and chunk_length") |
| return ConstantPolicy(**parameters) |
|
|
