Crash

At a glance

The community member is trying to use the BM25Retriever with Chroma, following the documentation, but their Jupyter Lab kernel keeps dying and restarting when they try to execute the VectorStoreIndex line. They are using a small PDF, so they don't think it's a memory overflow issue. The other community members suggest that the issue might be on Chroma's side, as they have faced similar issues with Chroma's stability and performance. One community member recommends switching to Qdrant instead of Chroma.

Useful resources

llucawang_nfls

I'm trying to use BM25Retriever w/ Chroma and this is the documentation that I follow: https://docs.llamaindex.ai/en/stable/examples/retrievers/bm25_retriever/#hybrid-retriever-with-bm25-chroma
When I try to execute this line: index = VectorStoreIndex(nodes=nodes, storage_context=storage_context), my jupyter lab kernel always dies and restarts.
I'm only using a very small PDF so there shouldn't be memory overflow issues.
Do you have any ideas of what's going on? Thanks!

Attachment

6 comments

LLogan M

Are you sure its small? 😅 what is len(nodes) ?

llucawang_nfls

print(len(nodes)) = 363
print(sum(len(str(node)) for node in nodes)) = 62013
This is just from 1 PDF w/ 13 pages

llucawang_nfls

it has 363 nodes because I'm using the Sentence-Window Retrieval, which basically takes each sentence as a node unit (and its surrounding 3 sentences as context, recorded in metadata)

ttitus

oh i faced this too... change vectorDB to Qdrant.

I've been facing a lot of issues with ChromaDB whether it's data ingestion, or vector search. I think their latest releases aren't stable

ttitus

it's chroma's side.

llucawang_nfls

ah that's unfortunate...thanks for letting me know!

Add a reply

Find answers from the community

Crash