def evaluator_master_fn()

in ml/eval/evaluation_pipeline.py [0:0]

• 11 lines of code
• 3 McCabe index (conditional complexity)

def evaluator_master_fn(eval_dataset: list[dict], 
                        reward_output_filepath: str, 
                        all_responses: str, 
                        language: str, 
                        new_model,
                        old_model="CohereForAI/aya-expanse-8b"): 
    # language is string for now, will be an object later with FeeLLanguage class definition with specific lanugage 
    # functionalities (will also store latest model and be much easier to handle such functions) 
    
    # 1. Reward score evaluation: 
    args = EvalArguments(bfloat16=True, 
                         reward_output_fmt='1-0',