langchain: create_extraction_chain does not work with other LLMs? Replicate models fails to load

🤖

Hello,

Thank you for reaching out and providing detailed information about the issue you’re facing.

The create_extraction_chain function is designed to work with specific language learning models (LLMs) and it seems like the Replicate model you’re trying to use might not be fully compatible with it. The function might be using specific methods or properties that are only available in certain LLMs, and not in the Replicate model.

As for the Replicate model failing to load, it could be due to a missing or incorrect REPLICATE_API_TOKEN. The Replicate model requires the REPLICATE_API_TOKEN environment variable to be set with your API token. If the token is missing or incorrect, the model will fail to load and result in an OutputParserException.

Another possible reason for the Replicate model failing to load could be the absence of the replicate Python package. The Replicate model requires the replicate Python package to be installed. If it’s not installed, an ImportError will be raised.

Please ensure that the REPLICATE_API_TOKEN is correctly set and the replicate Python package is installed. You can install the replicate package using pip:

pip install replicate

If you’ve already done these steps and are still facing the issue, it would be helpful if you could provide more information about your environment, such as the Python version you’re using and the list of installed packages (pip freeze).

I hope this helps! Let me know if you have any other questions or if there’s anything else I can do to assist you.

Best, Dosu

Sources

• libs/langchain/langchain/llms/replicate.py
• libs/langchain/langchain/chains/llm.py
• libs/langchain/langchain/chains/openai_functions/extraction.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.

Has anyone solved this issue?

Year 2023 Same Question.

Code

repo_id="mistralai/Mistral-7B-Instruct-v0.1"

    llm = HuggingFaceHub(repo_id=repo_id)

    class Properties(BaseModel):
        person_name: str
        person_height: int
        person_hair_color: str
        dog_breed: Optional[str]
        dog_name: Optional[str]

    schema = {
    "properties": {
        "person_name": {"type": "string"},
        "person_height": {"type": "integer"},
        "person_hair_color": {"type": "string"},
        "dog_name": {"type": "string"},
        "dog_breed": {"type": "string"},
        "dog_extra_info": {"type": "string"},
    },
}
chain = create_extraction_chain_pydantic(pydantic_schema=Properties, llm=llm)

inp = """Alex is 5 feet tall. Claudia is 1 feet taller Alex and jumps higher than him. Claudia is a brunette and Alex is blonde."""

chain.run(inp)

ISSUE

raise OutputParserException( langchain_core.exceptions.OutputParserException: This output parser can only be used with a chat generation.

Extraction use case docs have been revamped: https://python.langchain.com/docs/use_cases/extraction

Please see documentation there with recommendations on how to extraction with LangChain.

It was indeed the problem. I managed to make it work wit the Kor library, which is just another way to create output parsors.

Thanks a lot for your response & help.

Really appreciate it !