src/sal/models/reward_models.py [294:304]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
            **model_kwargs,
        ).eval()

        return model, tokenizer

    def score(
        self, questions: list[str], outputs: list[list[str]]
    ) -> list[list[float]]:
        # reference code: https://huggingface.co/Skywork/Skywork-o1-Open-PRM-Qwen-2.5-7B#huggingface-inference
        all_scores = []
        for question, answers in zip(questions, outputs):
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



src/sal/models/reward_models.py [342:352]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
            **model_kwargs,
        ).eval()

        return model, tokenizer

    def score(
        self, questions: list[str], outputs: list[list[str]]
    ) -> list[list[float]]:
        all_scores = []

        for question, answers in zip(questions, outputs):
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -