Function create_evaluation_request

RealChart2Code_eval/evaluate_task2.py:523–564

(difficulty: str, category: str, data_info: str, image_path:str,
                            image_base64: str, evaluation_prompt: str, model_name: str)

def create_evaluation_request(difficulty: str, category: str, data_info: str, image_path:str,
                            image_base64: str, evaluation_prompt: str, model_name: str) -> Dict:
    
    # task_prompt = f"""# Task
    # Category: {category}
    # """
    # Instruction: {instruction}
    text_prompt = create_user_prompt(category, data_info)
    image_chart = encode_image_to_base64(image_path)
    
    user_content = [
        {
            "type": "text",
            "text": f"**Task and Data**:\n{text_prompt}\n\n**Generated Chart**: Please evaluate based on the following generated chart image."
        },
        {
            "type": "image_url",
            "image_url": {
                "url": image_base64
            }
        },
        {
            "type": "text",
            "text": f"**Provided Chart**: Please evaluate based on the following provided chart image. "
        },
        {
            "type": "image_url",
            "image_url": {
                "url": image_chart
            }
        }
    ]

    return {
        "model": model_name,
        "messages": [
            {"role": "system", "content": evaluation_prompt},
            {"role": "user", "content": user_content}
        ],
        "max_tokens": 8192,
        "temperature": 0.0,
    }

def get_gt_code_path(image_path: Path) -> Path:
    """根据image_path获取对应的ground truth代码文件路径"""