Replies: 6 comments 5 replies
-
|
The exception "RunnableSequence' object has no attribute 'get'" when instantiating In the provided code, the Here is a snippet showing the correct usage: from langchain_core.callbacks import Callbacks
# Assuming you have a valid Callbacks object
callbacks = Callbacks()
reduce_documents_chain = ReduceDocumentsChain(
combine_documents_chain=combine_documents_chain,
collapse_documents_chain=combine_documents_chain,
token_max=4000,
verbose=True,
callbacks=callbacks, # Ensure this is a Callbacks object
)Make sure that the Additionally, ensure that the from langchain.chains import (
StuffDocumentsChain,
LLMChain,
ReduceDocumentsChain,
MapReduceDocumentsChain,
)
from langchain_core.prompts import PromptTemplate
from langchain_community.llms import OpenAI
document_prompt = PromptTemplate(
input_variables=["page_content"],
template="{page_content}"
)
document_variable_name = "context"
llm = OpenAI()
prompt = PromptTemplate.from_template(
"Summarize this content: {context}"
)
llm_chain = LLMChain(llm=llm, prompt=prompt)
reduce_prompt = PromptTemplate.from_template(
"Combine these summaries: {context}"
)
reduce_llm_chain = LLMChain(llm=llm, prompt=reduce_prompt)
combine_documents_chain = StuffDocumentsChain(
llm_chain=reduce_llm_chain,
document_prompt=document_prompt,
document_variable_name=document_variable_name
)
reduce_documents_chain = ReduceDocumentsChain(
combine_documents_chain=combine_documents_chain,
)
chain = MapReduceDocumentsChain(
llm_chain=llm_chain,
reduce_documents_chain=reduce_documents_chain,
)Ensure that all components are correctly instantiated and passed to avoid attribute errors [1][2]. |
Beta Was this translation helpful? Give feedback.
-
|
@dubbl-d please provide a minimal reproducible example. Feel free to open an issue if it's obviously not from user code |
Beta Was this translation helpful? Give feedback.
-
|
@eyurtsev Sure thing. Sorry the first pass didn't have a self contained example. Things fall over in the ctor for the ReduceDocumentsChain. Let me know if you need anything else. Note that this happens when using the def create_topic_summary():
from langchain.chains import ReduceDocumentsChain
from langchain.chains.combine_documents.stuff import create_stuff_documents_chain
from langchain_core.prompts import PromptTemplate
from langchain_openai import ChatOpenAI
conversation = """
Hi John. Hi Joe. Can I buy you a coffee? Sure. I'll have a cappuccino. By the way, I got that home espresso
machine you sent me a link to. We are enjoying it. Every Sunday we have a gourmet coffee experience with our breakfast.
We are trying out classic espresso drinks from across Europe. It is a lot of fun. That's great. That is a
fun thing to do. Yeah. We'll have you guys over for a brunch on Sunday and take a request for your drink of
choice. We would love that. Did you get to those trade shows last quarter? Yes. I went to all of the large to
mid sized shows this season. Did you get a chance to make some C-Level executive introductions? I did. I have
been able to set up product demonstrations with the primary 200-500 employee companies in the industry. So,
we are working on making those connections deeper and also targeting them with ads to reinforce our brand
recognition.
"""
topic_summary_map_prompt = """Analyze the conversation and determine the topics that were discussed. Create a
list of topics and for each topic include a short summary of the conversation around that topic.
{conversation}
"""
topic_summary_reduce_prompt = """The following is a set of topics and topic summaries:
{topic_summaries}
Take these summaries and distill them into a final, consolidated summary of the conversation.
"""
llm = ChatOpenAI(openai_api_key='USE-YOUR-OPENAI-KEY',
model='gpt-4-turbo', temperature=0)
map_prompt = PromptTemplate(template=topic_summary_map_prompt,
input_variables=["conversation"])
map_chain = map_prompt | llm
reduce_prompt = PromptTemplate.from_template(template=topic_summary_reduce_prompt)
combine_documents_chain = create_stuff_documents_chain(llm, reduce_prompt,
document_variable_name="topic_summaries")
reduce_documents_chain = ReduceDocumentsChain(combine_documents_chain=combine_documents_chain,
collapse_documents_chain=combine_documents_chain,
token_max=4000, verbose=True) |
Beta Was this translation helpful? Give feedback.
-
|
OK bug caused by pydantic being stricter with validation -- we need to update the type information in ReduceDocumentChain |
Beta Was this translation helpful? Give feedback.
-
|
I encountered the same issue when setting Stacktrace: |
Beta Was this translation helpful? Give feedback.
-
|
Likely due to the same issue as the failure above |
Beta Was this translation helpful? Give feedback.
-
|
Upgraded to issue #26840 |
Beta Was this translation helpful? Give feedback.
-
|
@dubbl-d did this work for you pre-0.3? I'm not sure that We are deprecating ReduceDocumentsChain in #26853. There's a recommended replacement implementation in this guide: https://python.langchain.com/docs/versions/migrating_chains/map_reduce_chain/ |
Beta Was this translation helpful? Give feedback.
-
|
@ccurme Before I upgraded from 0.1 to 0.3 it worked only because I was still using the In this new recommended method, can I still get to the intermediate steps? Getting access to those is important as it lets me get access to the output of each chunk which has an important use in my use case. Thanks in advance for that. |
Beta Was this translation helpful? Give feedback.
-
|
@dubbl-d, following will avoid the depcreation error and will achieve the same functionality that you need. I am not sure what you mean by llm = ChatOpenAI()
map_prompt = "Summarize the following doc: {context}"
map_prompt = PromptTemplate.from_template(template=map_prompt)
reduce_prompt = PromptTemplate.from_template("""
The following is a set of summaries:
{docs}
Take these and distill it into a final, consolidated summary
of the main themes.
""")
map_chain = create_stuff_documents_chain(llm, map_prompt, document_variable_name="context")
reduce_chain = reduce_prompt | llm | StrOutputParser()
class OverallState(TypedDict):
contents: List[str]
summaries: Annotated[list, operator.add]
final_summary: str
class SummaryState(TypedDict):
content: str
def generate_summary(state: SummaryState):
response = map_chain.invoke({"context":[state["content"]]})
return {"summaries": [response]}
def map_summaries(state: OverallState):
return [
Send("generate_summary", {"content": content}) for content in state["contents"]
]
def generate_final_summary(state: OverallState):
response = reduce_chain.invoke(state["summaries"])
return {"final_summary": response}
graph = StateGraph(OverallState)
graph.add_node("generate_summary", generate_summary)
graph.add_node("generate_final_summary", generate_final_summary)
graph.add_conditional_edges(START, map_summaries, ["generate_summary"])
graph.add_edge("generate_summary", "generate_final_summary")
graph.add_edge("generate_final_summary", END)
app = graph.compile()
print(app.invoke({"contents": [Document(page_content = "Langchain is amazing"),
Document(page_content = "Langchain just raised some more funding")]}))
|
Beta Was this translation helpful? Give feedback.
-
|
@keenborder786 Thanks. Yeah I found this on a migration guide, but I appreciate the pointer to it in case it had eluded me. Good stuff. In the MapReduceChain the In any case, I appreciate the helpful and thoughtful reply to my post. Thank you mate! |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Checked other resources
Commit to Help
Example Code
Description
I am trying to get rid of deprecation warnings and one of the things I have to do is replace the StuffDocumentsChain with the preferred
create_stuff_documents_chain(...)method. In the code above, I am getting an exception in pydantic validation code that is looking for agetattribute on theRunnableSequenceobject. I have checked with the 0.3 langchain docs to make sure my dependencies are correct and I don't reference any Pydantic v1 objects anywhere. This exception happens in theReduceDocumentsChainconstructor/init method when it calls pydantic to validate attributes. It isn't clear where or why it is looking to validate agetattribute.My dependencies and system info are in the section below. Here is the stack trace that I am getting from LangSmith traces:
AttributeError("'RunnableSequence' object has no attribute 'get'") Traceback (most recent call last): File "/Users/dubbled/eng/inference/llm/llm_inferences.py", line 108, in infer_transcript_topic_summary reduce_documents_chain = ReduceDocumentsChain(combine_documents_chain=combine_documents_chain, ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/Users/dubbled/.pyenv/versions/3.11.7/lib/python3.11/site-packages/langchain_core/load/serializable.py", line 112, in init super().init(*args, **kwargs) File "/Users/dubbled/.pyenv/versions/3.11.7/lib/python3.11/site-packages/pydantic/main.py", line 212, in init validated_self = self.pydantic_validator.validate_python(data, self_instance=self) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/Users/dubbled/.pyenv/versions/3.11.7/lib/python3.11/site-packages/langchain/chains/base.py", line 236, in raise_callback_manager_deprecation if values.get("callback_manager") is not None: ^^^^^^^^^^ File "/Users/dubbled/.pyenv/versions/3.11.7/lib/python3.11/site-packages/langchain_core/runnables/base.py", line 5704, in getattr attr = getattr(self.bound, name) ^^^^^^^^^^^^^^^^^^^^^^^^^ File "/Users/dubbled/.pyenv/versions/3.11.7/lib/python3.11/site-packages/pydantic/main.py", line 856, in getattr raise AttributeError(f'{type(self).name!r} object has no attribute {item!r}') AttributeError: 'RunnableSequence' object has no attribute 'get'
System Info
$ python -m langchain_core.sys_info
System Information
Package Information
Optional packages not installed
Other Dependencies
Beta Was this translation helpful? Give feedback.
All reactions