Method evaluate

src/evaluation/evaluator.py:41–54

(
        self,
        *,
        trajectories_path: Path,
        scenarios_paths: list[Path],
    )

        self.judge_model = judge_model

    def evaluate(
        self,
        *,
        trajectories_path: Path,
        scenarios_paths: list[Path],
    ) -> EvalReport:
        scenarios = load_scenarios(scenarios_paths)
        trajectories = load_trajectories(trajectories_path)

        results: list[ScenarioResult] = []
        for scenario, traj in join_records(scenarios, trajectories):
            results.append(self._score_one(scenario, traj))

        return build_report(results)

    def _score_one(
        self, scenario: Scenario, traj: PersistedTrajectory