An issue in Langchain when using it for summarization

[balachander1964]

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

#Imports
import os, sys, datetime, path
import pypdf
import tracemalloc
from langchain import chains
import openai
from openai import ChatCompletion 
from azure.cosmos import CosmosClient, PartitionKey
from langchain.chains import summarize, retrieval_qa
from langchain.chains.summarize import load_summarize_chain, map_reduce_prompt, refine_prompts
from langchain.chains import mapreduce 
from langchain.chains.combine_documents.map_reduce import MapReduceDocumentsChain
from langchain.chains.combine_documents.reduce import ReduceDocumentsChain
from langchain.chains.combine_documents.stuff import StuffDocumentsChain
from langchain.text_splitter import NLTKTextSplitter, CharacterTextSplitter
from langchain.chains.llm import LLMChain
from langchain_core.prompts import PromptTemplate
from langchain_core.output_parsers import StrOutputParser
from langchain import hub
from langchain_community.document_loaders import PyPDFLoader, TextLoader

from langchain.embeddings import OpenAIEmbeddings
from langchain.vectorstores import AzureCosmosDBVectorSearch, azure_cosmos_db
from langchain.llms import OpenAI
from langchain_openai import ChatOpenAI


#Add your open AI key to the system variable and access it.
openai.api_key = os.environ.get('openai_key')
_MODEL = 'gpt-4o'
_LLM = ChatOpenAI(temperature=0, model_name= _MODEL)

_SRC_PATH = path.Path('d:/sunwaretech/resumes/Yuvaraj Res.pdf')


def main():
  '''Main Entry'''
  #Load the pdf and retrieve the document.
  loader = PyPDFLoader(_SRC_PATH)
  docs = loader.load_and_split()
  doc_splitter = CharacterTextSplitter.from_tiktoken_encoder(
    chunk_size=1024, chunk_overlap=0
    )

  pages = doc_splitter.split_documents(docs)
  
  # Map
  map_prompt_template = """Write a concise summary of the following chunk of text that includes the main points and any important details.
  {Text}
  Helpful Answer:"""
  
  # Reduce
  combine_prompt_template = """Write a concise summary of the following text, delimited by newline character.
  Return your response in bullet points which covers the key points of the text.
  {Text}
  Helpful Answer:"""

#Map Reduce code  
  input_vars = ['Text']
  map_prompt = PromptTemplate.from_template(template = map_prompt_template, input_variables = input_vars)
  #map_chain = LLMChain(llm=_LLM, prompt=map_prompt)
  map_chain = map_prompt | _LLM| StrOutputParser()

  combine_prompt = PromptTemplate.from_template( template = combine_prompt_template, input_variables=input_vars)

  #Run the Chain.
  combine_chain = combine_prompt | _LLM| StrOutputParser()

  # Generate summaries using MapReduce method
  #After  defining prompts, initialize the associated map_reduce_chain.
  map_reduce_chain = load_summarize_chain(
    llm = _LLM,
    chain_type="map_reduce",
    map_prompt=map_prompt,
    combine_prompt=combine_prompt,
    document_variable_name = 'pages',
    return_intermediate_steps=True,
    )

  #Then, you generate summaries using the chain. Notice that LangChain use a tokenizer (from transformer library) with 1024 token limit by default.
  map_reduce_outputs  = map_reduce_chain({"input_documents": pages})
  print(map_reduce_outputs )

  return

'''
  Required for all python programs.
'''
if __name__ == '__main__':
  tracemalloc.start()
  print('Starting')
  main()
  tracemalloc.stop()
  
  input('Done')

Description

Hi, I got the following error when using Langchain for Summarising a PDF document. I will appreciate if you let me know how to fix this.
Traceback (most recent call last):
File "D:\Projects2017\langchain_gpt_01\langchain_gpt_01\summarizer_qa.py", line 92, in <module>
main()
File "D:\Projects2017\langchain_gpt_01\langchain_gpt_01\summarizer_qa.py", line 60, in main
map_prompt = PromptTemplate.from_template(template = map_prompt_template, input_variables = input_vars)
File "D:\python\Anaconda3\envs\spacy_-langchain\lib\site-packages\langchain_core\prompts\prompt.py", line 269, in from_template
return cls(
TypeError: ModelMetaclass object got multiple values for keyword argument 'input_variables'

System Info

Editor : Microsoft Visual Studio 2022, Python Version : 3.8x, OS : Windows 10 Pro

[dosubot[bot]]

Hello, @balachander1964! I'm here to assist you with any issues or questions you have. Let's get started!

The error you're encountering, TypeError: ModelMetaclass object got multiple values for keyword argument 'input_variables', typically occurs when the input_variables argument is passed multiple times when creating instances of PromptTemplate.

To resolve this, ensure that you are not passing input_variables more than once. Here is the corrected version of your code:

# Imports
import os, sys, datetime, path
import pypdf
import tracemalloc
from langchain import chains
import openai
from openai import ChatCompletion 
from azure.cosmos import CosmosClient, PartitionKey
from langchain.chains import summarize, retrieval_qa
from langchain.chains.summarize import load_summarize_chain, map_reduce_prompt, refine_prompts
from langchain.chains import mapreduce 
from langchain.chains.combine_documents.map_reduce import MapReduceDocumentsChain
from langchain.chains.combine_documents.reduce import ReduceDocumentsChain
from langchain.chains.combine_documents.stuff import StuffDocumentsChain
from langchain.text_splitter import NLTKTextSplitter, CharacterTextSplitter
from langchain.chains.llm import LLMChain
from langchain_core.prompts import PromptTemplate
from langchain_core.output_parsers import StrOutputParser
from langchain import hub
from langchain_community.document_loaders import PyPDFLoader, TextLoader

from langchain.embeddings import OpenAIEmbeddings
from langchain.vectorstores import AzureCosmosDBVectorSearch, azure_cosmos_db
from langchain.llms import OpenAI
from langchain_openai import ChatOpenAI

# Add your open AI key to the system variable and access it.
openai.api_key = os.environ.get('openai_key')
_MODEL = 'gpt-4o'
_LLM = ChatOpenAI(temperature=0, model_name=_MODEL)

_SRC_PATH = path.Path('d:/sunwaretech/resumes/Yuvaraj Res.pdf')

def main():
    '''Main Entry'''
    # Load the pdf and retrieve the document.
    loader = PyPDFLoader(_SRC_PATH)
    docs = loader.load_and_split()
    doc_splitter = CharacterTextSplitter.from_tiktoken_encoder(
        chunk_size=1024, chunk_overlap=0
    )

    pages = doc_splitter.split_documents(docs)
  
    # Map
    map_prompt_template = """Write a concise summary of the following chunk of text that includes the main points and any important details.
    {Text}
    Helpful Answer:"""
  
    # Reduce
    combine_prompt_template = """Write a concise summary of the following text, delimited by newline character.
    Return your response in bullet points which covers the key points of the text.
    {Text}
    Helpful Answer:"""

    # Map Reduce code  
    input_vars = ['Text']
    map_prompt = PromptTemplate(template=map_prompt_template, input_variables=input_vars)
    map_chain = map_prompt | _LLM | StrOutputParser()

    combine_prompt = PromptTemplate(template=combine_prompt_template, input_variables=input_vars)
    combine_chain = combine_prompt | _LLM | StrOutputParser()

    # Generate summaries using MapReduce method
    map_reduce_chain = load_summarize_chain(
        llm=_LLM,
        chain_type="map_reduce",
        map_prompt=map_prompt,
        combine_prompt=combine_prompt,
        document_variable_name='pages',
        return_intermediate_steps=True,
    )

    map_reduce_outputs = map_reduce_chain({"input_documents": pages})
    print(map_reduce_outputs)

    return

if __name__ == '__main__':
    tracemalloc.start()
    print('Starting')
    main()
    tracemalloc.stop()
    input('Done')

Ensure that you are not passing input_variables more than once when creating PromptTemplate instances. This should resolve the "TypeError" you are encountering ^[1].

To continue talking to Dosu, mention @dosu.