TypeError: Expected a Runnable, callable or dict.Instead got an unsupported type: <class 'list'>

Question

I am experimenting with LLM development.

Here is my code:

import langchain, pydantic, transformers
from langchain import HuggingFacePipeline
from langchain_core.prompts import PromptTemplate
from langchain_core.runnables.base import RunnableSequence, RunnableMap, RunnableLambda

from langchain.callbacks import get_openai_callback
from pydantic import BaseModel, Field
from langchain.output_parsers import PydanticOutputParser
from transformers import pipeline

class MedicalSpecialty(BaseModel):
    medical_specialty: str = Field(description="medical specialty the patient should go to")
    urgent: bool = Field(description="the patient should go to the hospital immediately")

parser = PydanticOutputParser(pydantic_object=MedicalSpecialty)

queries = ["i have ache in my chest and in my left arm. Which medical specialty should I go to?"]

template = """
Question: {question}
"""
prompt = PromptTemplate(template=template, input_variables=["question"])


llm = HuggingFacePipeline.from_model_id(
    model_id="bigscience/bloom-1b7",
    task="text-generation",
    model_kwargs={"max_length": 1024},
    device=-1  # Ensure it runs on CPU for macOS M1
)

# Wrap the prompt in a RunnableLambda to make it a Runnable
prompt_runnable = RunnableLambda(lambda x: prompt.format(**x))

# Define the sequence that includes the prompt and LLM
sequence = RunnableSequence([
    prompt_runnable,
    llm
])

with get_openai_callback() as CB:
    for query in queries:
        result = sequence.invoke({"question": query})
        print(query)
        print(result)
        print("====================================")
    
    # Print the costs of the requests
    print(cb)

Unfortunately, after several iterations, I keep getting this error:

TypeError                                 Traceback (most recent call last)
Cell In[6], line 19
     16 prompt_runnable = RunnableLambda(lambda x: prompt.format(**x))
     18 # Define the sequence that includes the prompt and LLM
---> 19 sequence = RunnableSequence([
     20     prompt_runnable,
     21     llm
     22 ])
     24 with get_openai_callback() as CB:
     25     for query in queries:

File /opt/anaconda3/envs/LLM/lib/python3.11/site-    packages/langchain_core/runnables/base.py:2632, in RunnableSequence.__init__(self,     name, first, middle, last, *steps)
   2630         steps_flat.extend(step.steps)
   2631     else:
-> 2632         steps_flat.append(coerce_to_runnable(step))
   2633 if len(steps_flat) < 2:
   2634     raise ValueError(
   2635         f"RunnableSequence must have at least 2 steps, got     {len(steps_flat)}"
   2636     )

File /opt/anaconda3/envs/LLM/lib/python3.11/site-    packages/langchain_core/runnables/base.py:5554, in coerce_to_runnable(thing)
   5552     return cast(Runnable[Input, Output], RunnableParallel(thing))
   5553 else:
-> 5554     raise TypeError(
   5555         f"Expected a Runnable, callable or dict."
   5556         f"Instead got an unsupported type: {type(thing)}"
   5557     )

TypeError: Expected a Runnable, callable or dict.Instead got an unsupported type:     <class 'list'>

Please, someone help!

Answer 1

No need to use lambda for this simple prompt with one input.

Try to use below and report back here on what happens

prompt_runnable = PromptRunnable(prompt)