Using chatGLM3-6b, the output result of create_sql_query_chain is incorrect.

[CharlieZZss]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

from langchain_community.utilities import SQLDatabase
from langchain.chains import create_sql_query_chain
import pymysql
db = SQLDatabase.from_uri("mysql+pymysql://root:root@localhost/zstudents");
llm = ChatGLM_LLM(model_path = "/root/autodl-tmp/ZhipuAI/chatglm3-6b")
chain = create_sql_query_chain(llm, db)
response = chain.invoke({"question": "How many students are there"})
response

Description

The issue originates from the documentation : https://python.langchain.com/docs/use_cases/sql/quickstart



When using ChatGLM3-6 as LLM, the query results generated by "create_sql_query_chain" are not in the format similar to the SQL statement "SELECT COUNT() FROM students," but rather in the format "'sql\nSELECT COUNT() FROM students;\n\nSQLResult: 3\nAnswer: 3'." How can this issue be resolved?I hope the result of its execution is an SQL Query.
I followed the tutorial to deploy the chatGLM3-6b model locally and customized an LLM class to integrate ChatGLM into the LangChain framework.https://github.com/datawhalechina/self-llm/blob/master/ChatGLM/05-ChatGLM3-6B%E6%8E%A5%E5%85%A5LangChain%E6%90%AD%E5%BB%BA%E7%9F%A5%E8%AF%86%E5%BA%93%E5%8A%A9%E6%89%8B.md

System Info

PyTorch 2.0.0
Python 3.8(ubuntu20.04)
pip install langchain==0.0.292

[dosubot[bot]]

🤖

Hey @CharlieZZss! Great to see you diving into the depths of LangChain again. How's the coding journey treating you this time?

Based on your request, it seems you want the create_sql_query_chain function to return an SQL query instead of the current format when using the ChatGLM3-6b model.

To achieve this, you can modify the create_sql_query_chain function to stop at the "SQLQuery:" token instead of the "SQLResult:" token. This will make the language model stop generating output after it has generated the SQL query, but before it generates the SQL result.

Here's how you can modify the function:

def create_sql_query_chain(
    llm: BaseLanguageModel,
    db: SQLDatabase,
    prompt: Optional[BasePromptTemplate] = None,
    k: int = 5,
) -> Runnable[Union[SQLInput, SQLInputWithTables, Dict[str, Any]], str]:
    ...
    return (
        RunnablePassthrough.assign(**inputs)  # type: ignore
        | (
            lambda x: {
                k: v
                for k, v in x.items()
                if k not in ("question", "table_names_to_use")
            }
        )
        | prompt_to_use.partial(top_k=str(k))
        | llm.bind(stop=["\nSQLQuery:"])  # Change this line
        | StrOutputParser()
        | _strip
    )

In the original function, the line llm.bind(stop=["\nSQLResult:"]) is used to make the language model stop generating output after it has generated the SQL result. By changing this line to llm.bind(stop=["\nSQLQuery:"]), you make the language model stop generating output after it has generated the SQL query.

Please note that this modification will only affect the output of the create_sql_query_chain function and not the actual execution of the SQL query. The SQL query will still be executed as before, but the output returned by the function will be the SQL query instead of the SQL result.

You can find the original function in the libs/langchain/langchain/chains/sql_database/query.py file in the LangChain repository.

I hope this helps! If you have any other questions or need further clarification, feel free to ask.

Sources

• libs/langchain/langchain/chains/sql_database/query.py

---

This response is meant to be useful and save you time. It is not meant to be a precise solution, but rather a starting point for your own research.

Help me be more useful! Please leave a 👍 if this is helpful and 👎 if it is irrelevant.

If you want to continue the conversation, start your reply with @dosu-bot.

[RollsChris]

Hi i'm having the same issue and i cant modify create_sql_query_chain? shouldn't this already be handled