src/fmeval/eval_algorithms/qa_accuracy.py [348:374]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        dataset_config: Optional[Union[DataConfig, List[DataConfig]]] = None,
        prompt_template: Optional[str] = None,
        num_records: int = 100,
        save: bool = False,
        save_strategy: Optional[SaveStrategy] = None,
    ) -> List[EvalOutput]:
        """Compute QA accuracy metrics on one or more datasets.

        :param model: An instance of ModelRunner representing the model under evaluation.
            If this argument is None, the `dataset_config` argument must not be None,
            and must correspond to a dataset that already contains a column with model outputs.
        :param dataset_config: Configures a single dataset or list of datasets used for the
            evaluation. If not provided, this method will run evaluations using all of its
            supported built-in datasets.
        :param prompt_template: A template used to generate prompts that are fed to the model.
            If not provided, defaults will be used. If provided, `model` must not be None.
        :param num_records: The number of records to be sampled randomly from the input dataset(s)
            used to perform the evaluation(s).
        :param save: If set to true, prompt responses and scores will be saved to a file.
        :param save_strategy: Specifies the strategy to use the save the localized outputs of the evaluations. If not
            specified, it will save it to the path that can be configured by the EVAL_RESULTS_PATH environment variable.
            If that environment variable is also not configured, it will be saved to the default path `/tmp/eval_results/`.

        :return: A list of EvalOutput objects.
        """
        # Create a shared resource to be used during the evaluation.
        bertscore_shared_resource = create_shared_resource(self.bertscore_model)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



src/fmeval/eval_algorithms/qa_accuracy_semantic_robustness.py [268:295]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        dataset_config: Optional[Union[DataConfig, List[DataConfig]]] = None,
        prompt_template: Optional[str] = None,
        num_records: int = 100,
        save: bool = False,
        save_strategy: Optional[SaveStrategy] = None,
    ) -> List[EvalOutput]:
        """Compute QA accuracy semantic robustness metrics on one or more datasets.

        :param model: An instance of ModelRunner representing the model under evaluation.
            This is a required argument, as even if the dataset contains model outputs,
            semantic robustness algorithms rely on invoking a model on perturbed inputs
            to see how the model outputs from the perturbed inputs differ from the original
            model outputs.
        :param dataset_config: Configures a single dataset or list of datasets used for the
            evaluation. If not provided, this method will run evaluations using all of its
            supported built-in datasets.
        :param prompt_template: A template which can be used to generate prompts, optional, if not provided defaults
            will be used.
        :param num_records: The number of records to be sampled randomly from the input dataset to perform the
                            evaluation
        :param save: If set to true, prompt responses and scores will be saved to a file.
        :param save_strategy: Specifies the strategy to use the save the localized outputs of the evaluations. If not
            specified, it will save it to the path that can be configured by the EVAL_RESULTS_PATH environment variable.
            If that environment variable is also not configured, it will be saved to the default path `/tmp/eval_results/`.
        :returns: A List of EvalOutput objects.
        """
        # Create a shared resource to be used during the evaluation.
        bertscore_shared_resource = create_shared_resource(self.bertscore_model)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -