How to go with HuggingFace* instead of ChatOpenAI?

[vriez]

Hi @tomasonjo , thank you very much for sharing this very informative material.

On this notebook, how could I change

llm = ChatOpenAI(model="gpt-3.5-turbo-16k", temperature=0)

to

from langchain.llms import HuggingFaceHub

llm = HuggingFaceHub(
    repo_id=repo_id, model_kwargs={"temperature": TEMPERATURE, "max_length": MAX_TOKENS}
)

or any other HuggingFacePipeline, and still make the tutorial work?

Of course, cypher_chain's llms would also have to be changed to other pipelines, but I have not got there yet.

The error I get is:

File ~/Projects/blogs/openaifunction_constructing_graph.py:277, in extract_and_store_graph(document, nodes, rels)
    274 
    275 extract_chain = get_extraction_chain(nodes, rels)
--> 277 data = extract_chain.run(document.page_content)
    278

File ~/anaconda3/envs/master/lib/python3.8/site-packages/langchain/chains/base.py:507, in Chain.run(self, callbacks, tags, metadata, *args, **kwargs)
    505     if len(args) != 1:
    506         raise ValueError("`run` supports only one positional argument.")
--> 507     return self(args[0], callbacks=callbacks, tags=tags, metadata=metadata)[
    508         _output_key
    509     ]
    511 if kwargs and not args:
    512     return self(kwargs, callbacks=callbacks, tags=tags, metadata=metadata)[
    513         _output_key
    514     ]

File ~/anaconda3/envs/master/lib/python3.8/site-packages/langchain/chains/base.py:312, in Chain.__call__(self, inputs, return_only_outputs, callbacks, tags, metadata, run_name, include_run_info)
    310 except BaseException as e:
    311     run_manager.on_chain_error(e)
--> 312     raise e
    313 run_manager.on_chain_end(outputs)
    314 final_outputs: Dict[str, Any] = self.prep_outputs(
    315     inputs, outputs, return_only_outputs
    316 )

File ~/anaconda3/envs/master/lib/python3.8/site-packages/langchain/chains/base.py:306, in Chain.__call__(self, inputs, return_only_outputs, callbacks, tags, metadata, run_name, include_run_info)
    299 run_manager = callback_manager.on_chain_start(
    300     dumpd(self),
    301     inputs,
    302     name=run_name,
    303 )
    304 try:
    305     outputs = (
--> 306         self._call(inputs, run_manager=run_manager)
    307         if new_arg_supported
    308         else self._call(inputs)
    309     )
    310 except BaseException as e:
    311     run_manager.on_chain_error(e)

File ~/anaconda3/envs/master/lib/python3.8/site-packages/langchain/chains/llm.py:104, in LLMChain._call(self, inputs, run_manager)
     98 def _call(
     99     self,
    100     inputs: Dict[str, Any],
    101     run_manager: Optional[CallbackManagerForChainRun] = None,
    102 ) -> Dict[str, str]:
    103     response = self.generate([inputs], run_manager=run_manager)
--> 104     return self.create_outputs(response)[0]

File ~/anaconda3/envs/master/lib/python3.8/site-packages/langchain/chains/llm.py:258, in LLMChain.create_outputs(self, llm_result)
    256 def create_outputs(self, llm_result: LLMResult) -> List[Dict[str, Any]]:
    257     """Create outputs from response."""
--> 258     result = [
    259         # Get the text of the top generated string.
    260         {
    261             self.output_key: self.output_parser.parse_result(generation),
    262             "full_generation": generation,
    263         }
    264         for generation in llm_result.generations
    265     ]
    266     if self.return_final_only:
    267         result = [{self.output_key: r[self.output_key]} for r in result]

File ~/anaconda3/envs/master/lib/python3.8/site-packages/langchain/chains/llm.py:261, in <listcomp>(.0)
    256 def create_outputs(self, llm_result: LLMResult) -> List[Dict[str, Any]]:
    257     """Create outputs from response."""
    258     result = [
    259         # Get the text of the top generated string.
    260         {
--> 261             self.output_key: self.output_parser.parse_result(generation),
    262             "full_generation": generation,
    263         }
    264         for generation in llm_result.generations
    265     ]
    266     if self.return_final_only:
    267         result = [{self.output_key: r[self.output_key]} for r in result]

File ~/anaconda3/envs/master/lib/python3.8/site-packages/langchain/output_parsers/openai_functions.py:174, in PydanticAttrOutputFunctionsParser.parse_result(self, result, partial)
    173 def parse_result(self, result: List[Generation], *, partial: bool = False) -> Any:
--> 174     result = super().parse_result(result)
    175     return getattr(result, self.attr_name)

File ~/anaconda3/envs/master/lib/python3.8/site-packages/langchain/output_parsers/openai_functions.py:157, in PydanticOutputFunctionsParser.parse_result(self, result, partial)
    156 def parse_result(self, result: List[Generation], *, partial: bool = False) -> Any:
--> 157     _result = super().parse_result(result)
    158     if self.args_only:
    159         pydantic_args = self.pydantic_schema.parse_raw(_result)  # type: ignore

File ~/anaconda3/envs/master/lib/python3.8/site-packages/langchain/output_parsers/openai_functions.py:26, in OutputFunctionsParser.parse_result(self, result, partial)
     24 generation = result[0]
     25 if not isinstance(generation, ChatGeneration):
---> 26     raise OutputParserException(
     27         "This output parser can only be used with a chat generation."
     28     )
     29 message = generation.message
     30 try:

OutputParserException: This output parser can only be used with a chat generation.

[GiacomoBelliniStudent]

I got stuck in the same error. Seems that code is not optimized for other llms yet

[tomasonjo]

It's not that code is not optimized for other LLMs, it's that most other LLMs are not optimized for Cypher generation