Kernal crash while excecuting vector embedding operation

[mehar-tejat]

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

# data ingestion from pdf
# pdf reader
# this will take each single page
from langchain_community.document_loaders import PyPDFLoader
loader = PyPDFLoader('hi.pdf')
docs = loader.load()

from langchain.text_splitter import RecursiveCharacterTextSplitter
text_splitter = RecursiveCharacterTextSplitter(chunk_size=10,chunk_overlap=1)
documents = text_splitter.split_documents(docs)
documents[:5]

# vector embedding and vector store
from langchain_community.embeddings import OllamaEmbeddings
from langchain_community.vectorstores import Chroma


# giving first 20 documents
db = Chroma.from_documents(documents[:1], OllamaEmbeddings())

Description

i am trying to build vector embedding using OllamaEmbeddings, as you can see in the image , while I'm trying to run this, i am getting "Kernel crash" I'm using environment python 3.10 .0 ,
i am getting in log "18:07:26.644 [error] Disposing session as kernel process died ExitCode: 3221225477 "
the error is showing in red as in below image
any help

System Info

data ingestion from pdf

pdf reader

this will take each single page

from langchain_community.document_loaders import PyPDFLoader
loader = PyPDFLoader('hi.pdf')
docs = loader.load()

from langchain.text_splitter import RecursiveCharacterTextSplitter
text_splitter = RecursiveCharacterTextSplitter(chunk_size=10,chunk_overlap=1)
documents = text_splitter.split_documents(docs)
documents[:5]

vector embedding and vector store

from langchain_community.embeddings import OllamaEmbeddings
from langchain_community.vectorstores import Chroma

giving first 20 documents

db = Chroma.from_documents(documents[:1], OllamaEmbeddings())

[dosubot[bot]]

@mehar-tejat Hello there! I'm here to help you with any bugs, questions, or contributions you may have.

The kernel crash during the execution of your vector embedding operation could be due to several factors, including deprecated classes, incorrect imports, or suboptimal configurations for your hardware. Here are some steps to help you resolve the issue:

1. Ensure Correct and Updated Imports:

   • Make sure you are using the correct and updated classes for embedding operations. For example, if you are using OllamaEmbeddings, ensure it is correctly imported.

2. Check for Deprecated Classes:

   • Verify that you are not using any deprecated classes. Update your code to use the recommended classes from the langchain_community.embeddings module.

3. Optimize for Your Hardware:

   • If you are using an Intel GPU, you can benefit from IPEX-LLM optimizations. Follow the setup instructions for your environment (Windows or Linux) to ensure optimal performance.

4. Enable Debugging:

   • Enable detailed logging to diagnose the issue. Use the set_debug function to print inputs and outputs of LangChain components.

Here is an example of how to enable debugging and ensure correct imports:

from langchain.globals import set_debug
set_debug(True)

from langchain_community.document_loaders import PyPDFLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.embeddings import OllamaEmbeddings
from langchain_community.vectorstores import Chroma

# Load PDF
loader = PyPDFLoader('hi.pdf')
docs = loader.load()

# Split text
text_splitter = RecursiveCharacterTextSplitter(chunk_size=10, chunk_overlap=1)
documents = text_splitter.split_documents(docs)

# Vector embedding and vector store
db = Chroma.from_documents(documents[:1], OllamaEmbeddings())

By following these steps, you can identify and resolve the cause of the kernel crash during your vector embedding operation ^[1][2][3].

To continue talking to Dosu, mention @dosu.

[Hmzbo]

Hello @mehar-tejat ,

I have faced this issue before when i was trying to run similar code on my PC (Windows 11, 16gb RAM). I found many talking about OOM error causing the kernel to crash, but that wasn't the case for me, I monitored my memory and tried with very small data and the issue persisted. I still don't know what's causing this.

However, I found a workaround that worked for me. I simply saved the ChromaDB on my disk and then load it to memory when computing similarity.

vectorstore = Chroma.from_documents(documents[:1], OllamaEmbeddings(), persist_directory="./chroma_db")

I have to mention that I used HuggingFaceEmbeddings instead of OllamaEmbeddings.

I hope this helps.

[mehar-tejat]

Hi @Hmzbo,

Thanks for the support, i tried for 2 days in multiple ways and found instead of Chroma db i have used FAISS db, it worked, i will try your approach.

[Amine-Fadssi]

Hello @Hmzbo,
Thanks for the tip! I tried saving the ChromaDB to disk and loading it into memory, and it worked. Appreciate the help!