1  import fire
 2  import pandas as pd
 3  from DABench import DABench
 4  
 5  from metagpt.logs import logger
 6  from metagpt.roles.di.data_interpreter import DataInterpreter
 7  from metagpt.utils.recovery_util import save_history
 8  
 9  
10  async def main():
11      """Evaluate all"""
12      bench = DABench()
13      id_list, predictions, labels, is_true = [], [], [], []
14      for key, value in bench.answers.items():
15          id_list.append(key)
16          labels.append(str(bench.get_answer(key)))
17          try:
18              requirement = bench.generate_formatted_prompt(key)
19              di = DataInterpreter()
20              result = await di.run(requirement)
21              logger.info(result)
22              save_history(role=di)
23              temp_prediction, temp_istrue = bench.eval(key, str(result))
24              is_true.append(str(temp_istrue))
25              predictions.append(str(temp_prediction))
26          except:
27              is_true.append(str(bench.eval(key, "")))
28              predictions.append(str(""))
29      df = pd.DataFrame({"Label": labels, "Prediction": predictions, "T/F": is_true})
30      df.to_excel("DABench_output.xlsx", index=False)
31      logger.info(bench.eval_all(id_list, predictions))
32  
33  
34  if __name__ == "__main__":
35      fire.Fire(main)