关于toolcall微调数据格式请教一下~

[Jimmy-L99]

Reminder

• I have read the README and searched the existing issues.

System Info

Reproduction

一条toolcall微调数据模板如下，来自于ChatGLM4：

{
  "messages": [
    {
      "role": "system",
      "content": "",
      "tools": [
        {
          "type": "function",
          "function": {
            "name": "get_recommended_books",
            "description": "Get recommended books based on user's interests",
            "parameters": {
              "type": "object",
              "properties": {
                "interests": {
                  "type": "array",
                  "items": {
                    "type": "string"
                  },
                  "description": "The interests to recommend books for"
                }
              },
              "required": [
                "interests"
              ]
            }
          }
        }
      ]
    },
    {
      "role": "user",
      "content": "Hi, I am looking for some book recommendations. I am interested in history and science fiction."
    },
    {
      "role": "assistant",
      "content": "{\"name\": \"get_recommended_books\", \"arguments\": {\"interests\": [\"history\", \"science fiction\"]}}"
    },
    {
      "role": "observation",
      "content": "{\"books\": [\"Sapiens: A Brief History of Humankind by Yuval Noah Harari\", \"A Brief History of Time by Stephen Hawking\", \"Dune by Frank Herbert\", \"The Martian by Andy Weir\"]}"
    },
    {
      "role": "assistant",
      "content": "Based on your interests in history and science fiction, I would recommend the following books: \"Sapiens: A Brief History of Humankind\" by Yuval Noah Harari, \"A Brief History of Time\" by Stephen Hawking, \"Dune\" by Frank Herbert, and \"The Martian\" by Andy Weir."
    }
  ]
}

想请教一下，如果按照上面模板构建数据，该如何填写dataset_info.json里面的tags：

  "glm4_toolcall_dataset":{
    "file_name": "",
    "formatting": "sharegpt",
    "columns": {
      "messages": "messages"
    },
    "tags":{
      "role_tag": "role",
      "content_tag": "content",
      "user_tag": "user",
      "assistant_tag": "assistant",
      "system_tag": "system",
    }
  }

1. 对话出现了两个assistant和一个observation，需要如何修改tags？

2. 还是说按照标准sharegpt重新构建数据？

3. 如果用langchain之类的框架，tools列表里面存在多个tool，传入llm的请求将会是：

2025-01-03 14:55:49.543 | INFO     | __main__:generate_stream_glm4:207 - inputs: [gMASK]<sop><|system|>
你是一个名为 GLM-4 的人工智能助手。你是基于智谱AI训练的语言模型 GLM-4 模型开发的，你的任务是针对用户的问题和要求提供适当的答复和支持。

# 可用工具
## add

{
    "name": "add",
    "description": "计算两个数字的和",
    "parameters": {
        "type": "object",
        "properties": {
            "a": {
                "description": "第一个数字",
                "type": "integer"
            },
            "b": {
                "description": "第二个数字",
                "type": "integer"
            }
        },
        "required": [
            "a",
            "b"
        ]
    }
}
在调用上述函数时，请使用 Json 格式表示调用的参数。
## multiply

{
    "name": "multiply",
    "description": "计算两个数字的乘积",
    "parameters": {
        "type": "object",
        "properties": {
            "a": {
                "description": "第一个数字",
                "type": "integer"
            },
            "b": {
                "description": "第二个数字",
                "type": "integer"
            }
        },
        "required": [
            "a",
            "b"
        ]
    }
}
在调用上述函数时，请使用 Json 格式表示调用的参数。
## weather

{
    "name": "weather",
    "description": "该工具只能得到指定城市的天气信息\n不能用于两个天气数据对比或其他用途",
    "parameters": {
        "type": "object",
        "properties": {
            "city": {
                "description": "城市名称",
                "type": "string"
            }
        },
        "required": [
            "city"
        ]
    }
}
在调用上述函数时，请使用 Json 格式表示调用的参数。
## purchase_order

{
    "name": "purchase_order",
    "description": "该工具只能获取文本中的订单的参数，\n分别是订单名称、采购物品、订单数量、订单价格",
    "parameters": {
        "type": "object",
        "properties": {
            "name": {
                "description": "订单名称",
                "type": "string"
            },
            "project": {
                "description": "采购物品",
                "type": "string"
            },
            "num": {
                "description": "订单数量",
                "type": "integer"
            },
            "price": {
                "description": "订单价格",
                "type": "integer"
            }
        },
        "required": [
            "name",
            "project",
            "num",
            "price"
        ]
    }
}
在调用上述函数时，请使用 Json 格式表示调用的参数。
## get_current_time

{
    "name": "get_current_time",
    "description": "该工具可以获取当前日期和时间",
    "parameters": {
        "type": "object",
        "properties": {
            "time": {
                "description": "时间",
                "type": "string"
            }
        },
        "required": [
            "time"
        ]
    }
}
在调用上述函数时，请使用 Json 格式表示调用的参数。<|user|>

构造标准sharegpt数据时，tools是要同时写进所有tool，还是只写入用到的某个tool？

问题有点长，且内容较多，还恳请解惑。

Expected behavior

希望更深入地了解toolcall微调的数据格式。

Others

No response