Function read_translation_dataset

evaluation/utils.py:257–287

(
    data_file_src: str = None,
    data_file_tgt: str = None,
    lang_src: str = None,
    lang_tgt: str = None,
    dataset_type: str = "humanevalx",
)

def read_translation_dataset(
    data_file_src: str = None,
    data_file_tgt: str = None,
    lang_src: str = None,
    lang_tgt: str = None,
    dataset_type: str = "humanevalx",
) -> Dict:
    if "humanevalx" in dataset_type.lower():
        dataset_src = {task["task_id"]: task for task in stream_jsonl(data_file_src)}
        dataset_tgt = {task["task_id"].split("/")[-1]: task for task in stream_jsonl(data_file_tgt)}
        for k, sample in dataset_src.items():
            prompt = "code translation\n"
            if lang_src == "cpp":
                prompt += "C++:\n"
            elif lang_src == "js":
                prompt += "JavaScript:\n"
            else:
                prompt += f"{lang_src}:\n".capitalize()
            prompt += dataset_src[k]["declaration"] + "\n" + dataset_src[k]["canonical_solution"].rstrip() + "\n"
            if lang_tgt == "cpp":
                prompt += "C++:\n"
            elif lang_tgt == "js":
                prompt += "JavaScript:\n"
            else:
                prompt += f"{lang_tgt}:\n".capitalize()
            prompt += dataset_tgt[k.split("/")[-1]]["declaration"]
            dataset_src[k]["prompt"] = prompt
    else:
        raise f"Dataset: {dataset_type} not supported."

    return dataset_src


def process_extra_prompt(