Method evaluate

AgentBench.old/src/task.py:84–100

(self, agent: Agent, already_runs: List[Any]=None)

        pass

    def evaluate(self, agent: Agent, already_runs: List[Any]=None) -> Dict[str, Any]:
        print(f"Evaluating task '{self.name}' ...")
        data = self.get_data()
        inputs = data.get_inputs()
        targets = data.get_targets()
        try:
            results = self.predict_all(agent, inputs, already_runs)
        except:
            results = self.predict_all(agent, inputs)
        # print(results)
        # print(type(results))
        result_dict = {}
        for metric in self.metrics:
            result_dict[metric] = self.metrics[metric](results, targets)
        print(f"Task '{self.name}' evaluation finished. The results are saved in '{self.get_output_dir()}'")
        self.save_runs_all(inputs, results, targets, result_dict)
        return result_dict

    def predict_all(self, agent: Agent, inputs: List[T_INPUT], already_runs: List[Any]=None) -> List[T_OUTPUT]:
        print(f"Start Predicting All ...")