[GH-ISSUE #11279] Embedding dimension 768 does not match collection dimensionality 384 #70772

Closed
opened 2026-05-13 02:47:32 -05:00 by GiteaMirror · 2 comments
Owner

Originally created by @rakehell1986 on GitHub (Mar 6, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/11279

2025-03-06 08:19:20.100 | INFO | open_webui.routers.retrieval:save_docs_to_vector_db:854 - collection 08ef5cae-3755-4bfa-8a23-2fe37afbf8d1 already exists - {}
2025-03-06 08:19:20.100 | INFO | open_webui.routers.retrieval:save_docs_to_vector_db:865 - adding to collection 08ef5cae-3755-4bfa-8a23-2fe37afbf8d1 - {}
2025-03-06 08:19:39.072 | ERROR | open_webui.routers.retrieval:save_docs_to_vector_db:904 - Embedding dimension 768 does not match collection dimensionality 384 - {}

after upgrading openwebui to 0.5.20,

i am using ollma embedding model : nomic-embed-text

any one know what should i config so that they are compatibal?

Originally created by @rakehell1986 on GitHub (Mar 6, 2025). Original GitHub issue: https://github.com/open-webui/open-webui/issues/11279 2025-03-06 08:19:20.100 | INFO | open_webui.routers.retrieval:save_docs_to_vector_db:854 - collection 08ef5cae-3755-4bfa-8a23-2fe37afbf8d1 already exists - {} 2025-03-06 08:19:20.100 | INFO | open_webui.routers.retrieval:save_docs_to_vector_db:865 - adding to collection 08ef5cae-3755-4bfa-8a23-2fe37afbf8d1 - {} 2025-03-06 08:19:39.072 | ERROR | open_webui.routers.retrieval:save_docs_to_vector_db:904 - Embedding dimension 768 does not match collection dimensionality 384 - {} after upgrading openwebui to 0.5.20, i am using ollma embedding model : [nomic-embed-text](https://ollama.com/library/nomic-embed-text) any one know what should i config so that they are compatibal?
Author
Owner

@rgaricano commented on GitHub (Mar 6, 2025):

the problem seem to be because you change the embedding model.
Each embedd model save info with a specific vector space, if model is changed probably it can't read this info.
If you remember what model was before change it, or if not you must to reset the vector storage database (at bottom of the same setting page)
(if you don't remember and you want to try, i think that all-minilm-l6 have 384 dimension, try with this model: sentence-transformers/all-MiniLM-L6-v2)

<!-- gh-comment-id:2704787015 --> @rgaricano commented on GitHub (Mar 6, 2025): the problem seem to be because you change the embedding model. Each embedd model save info with a specific vector space, if model is changed probably it can't read this info. If you remember what model was before change it, or if not you must to reset the vector storage database (at bottom of the same setting page) (if you don't remember and you want to try, i think that all-minilm-l6 have 384 dimension, try with this model: `sentence-transformers/all-MiniLM-L6-v2`)
Author
Owner

@j820301 commented on GitHub (Mar 7, 2025):

This is a problem I once encountered, and I got an effective solution. Please refer to it
If you update your embedding model you should reset/delete your existing knowledge.

#10153

<!-- gh-comment-id:2705317542 --> @j820301 commented on GitHub (Mar 7, 2025): This is a problem I once encountered, and I got an effective solution. Please refer to it If you update your embedding model you should reset/delete your existing knowledge. #10153
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/open-webui#70772