[GH-ISSUE #3606] multilingual-e5-large and multilingual-e5-base Embedding Model Support #64261

Open
opened 2026-05-03 16:49:10 -05:00 by GiteaMirror · 22 comments
Owner

Originally created by @awilhelm-projects on GitHub (Apr 11, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/3606

What are you trying to do?

I want use multilingual-e5-large or multilingual-e5-base as embedding model, because all other embed models dont work for other languages as english.

How should we solve this?

Convert multilingual-e5-large and multilingual-e5-base (https://huggingface.co/intfloat/multilingual-e5-base) to gguf and integrate it as models in ollama.

What is the impact of not solving this?

You cannot use ollama for RAG Application with languages other then english.

Anything else?

No response

Originally created by @awilhelm-projects on GitHub (Apr 11, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/3606 ### What are you trying to do? I want use multilingual-e5-large or multilingual-e5-base as embedding model, because all other embed models dont work for other languages as english. ### How should we solve this? Convert multilingual-e5-large and multilingual-e5-base (https://huggingface.co/intfloat/multilingual-e5-base) to gguf and integrate it as models in ollama. ### What is the impact of not solving this? You cannot use ollama for RAG Application with languages other then english. ### Anything else? _No response_
GiteaMirror added the model label 2026-05-03 16:49:11 -05:00
Author
Owner

@pksorensen commented on GitHub (Apr 19, 2024):

@awilhelm-projects - did you figure out how to convert multilingual-e5-large to gguf for llama.cpp ?

<!-- gh-comment-id:2067279251 --> @pksorensen commented on GitHub (Apr 19, 2024): @awilhelm-projects - did you figure out how to convert multilingual-e5-large to gguf for llama.cpp ?
Author
Owner

@awilhelm-projects commented on GitHub (Apr 24, 2024):

No - i have no idea. I think, it works with the tools of llama cpp (like for other models), but it seems llama cpp dont support xlm-roberta.

<!-- gh-comment-id:2075846202 --> @awilhelm-projects commented on GitHub (Apr 24, 2024): No - i have no idea. I think, it works with the tools of llama cpp (like for other models), but it seems llama cpp dont support xlm-roberta.
Author
Owner

@gevzak commented on GitHub (May 2, 2024):

+1
This is a much needed feature for me as well, because currently there are no multilingual embedding models I know of which are supported.

<!-- gh-comment-id:2089891991 --> @gevzak commented on GitHub (May 2, 2024): +1 This is a much needed feature for me as well, because currently there are no multilingual embedding models I know of which are supported.
Author
Owner

@dermatzeimnetz commented on GitHub (May 3, 2024):

+1
Seems there is currently no working multilingual gguf for embeddings only https://huggingface.co/models?sort=likes&search=multilingual+Gguf

<!-- gh-comment-id:2092928055 --> @dermatzeimnetz commented on GitHub (May 3, 2024): +1 Seems there is currently no working multilingual gguf for embeddings only https://huggingface.co/models?sort=likes&search=multilingual+Gguf
Author
Owner

@qsdhj commented on GitHub (May 14, 2024):

@awilhelm-projects
Hi,
I try to create a RAG application for german data. It would be great if we had a working multilingual embedding model as gguf.

Without batch processing of embeddings, ollama would still not be usable for RAG applications, or do you now how to enable this?

<!-- gh-comment-id:2109602552 --> @qsdhj commented on GitHub (May 14, 2024): @awilhelm-projects Hi, I try to create a RAG application for german data. It would be great if we had a working multilingual embedding model as gguf. Without batch processing of embeddings, ollama would still not be usable for RAG applications, or do you now how to enable this?
Author
Owner

@stygmate commented on GitHub (May 20, 2024):

same here for french

<!-- gh-comment-id:2119870992 --> @stygmate commented on GitHub (May 20, 2024): same here for french
Author
Owner

@beko-dt commented on GitHub (Jun 4, 2024):

+1
Would be great to have a multilingual model!

<!-- gh-comment-id:2146905184 --> @beko-dt commented on GitHub (Jun 4, 2024): +1 Would be great to have a multilingual model!
Author
Owner

@koakh commented on GitHub (Jun 11, 2024):

same here to Portuguese and other languages to

<!-- gh-comment-id:2160958838 --> @koakh commented on GitHub (Jun 11, 2024): same here to Portuguese and other languages to
Author
Owner

@KodeurKubik commented on GitHub (Jul 1, 2024):

Upvote for French :)

<!-- gh-comment-id:2200995205 --> @KodeurKubik commented on GitHub (Jul 1, 2024): Upvote for French :)
Author
Owner

@chikiu-san commented on GitHub (Jul 3, 2024):

Upvote for Japanese ;)

<!-- gh-comment-id:2205677730 --> @chikiu-san commented on GitHub (Jul 3, 2024): Upvote for Japanese ;)
Author
Owner

@MorrisLu-Taipei commented on GitHub (Jul 7, 2024):

Upvote for Chinese

<!-- gh-comment-id:2212538334 --> @MorrisLu-Taipei commented on GitHub (Jul 7, 2024): Upvote for Chinese
Author
Owner

@mdielacher commented on GitHub (Jul 12, 2024):

Upvote 4 multilingual-e5! :)

<!-- gh-comment-id:2225010672 --> @mdielacher commented on GitHub (Jul 12, 2024): Upvote 4 multilingual-e5! :)
Author
Owner

@celaltrk commented on GitHub (Jul 19, 2024):

Upvote for Turkish

<!-- gh-comment-id:2238312856 --> @celaltrk commented on GitHub (Jul 19, 2024): Upvote for Turkish
Author
Owner

@keix1 commented on GitHub (Jul 19, 2024):

Upvote!

<!-- gh-comment-id:2238480881 --> @keix1 commented on GitHub (Jul 19, 2024): Upvote!
Author
Owner

@dobry21 commented on GitHub (Jul 22, 2024):

Upvote 4 multilingual-e5! :)

<!-- gh-comment-id:2243716859 --> @dobry21 commented on GitHub (Jul 22, 2024): Upvote 4 multilingual-e5! :)
Author
Owner

@dobry21 commented on GitHub (Jul 23, 2024):

I have found out that this model is available in Local AI gallery https://localai.io/gallery.html under the name openvino-multilingual-e5-base

Perhaps this will be helpful for conversion to Ollama.

<!-- gh-comment-id:2244831997 --> @dobry21 commented on GitHub (Jul 23, 2024): I have found out that this model is available in Local AI gallery https://localai.io/gallery.html under the name openvino-multilingual-e5-base Perhaps this will be helpful for conversion to Ollama.
Author
Owner

@josegtmonteiro commented on GitHub (Jul 25, 2024):

+1

<!-- gh-comment-id:2250136662 --> @josegtmonteiro commented on GitHub (Jul 25, 2024): +1
Author
Owner

@u-masao commented on GitHub (Jul 25, 2024):

+1

I am also one of the people who tried to run Multilingual-E5 on Ollama.

A few months ago, I tried to create a customized model on Ollama, but was told that they do not support anything other than BPE.

So I'm serving the model with this simple code:

https://github.com/u-masao/sentence-transformers-cached-server/blob/main/app/main.py

<!-- gh-comment-id:2250377982 --> @u-masao commented on GitHub (Jul 25, 2024): +1 I am also one of the people who tried to run Multilingual-E5 on Ollama. A few months ago, I tried to create [a customized model](https://github.com/ollama/ollama?tab=readme-ov-file#import-from-pytorch-or-safetensors) on Ollama, but was told that they do not support anything other than BPE. So I'm serving the model with this simple code: https://github.com/u-masao/sentence-transformers-cached-server/blob/main/app/main.py
Author
Owner

@syakesaba commented on GitHub (Aug 15, 2024):

+1

Upvote for Japanese

<!-- gh-comment-id:2291160709 --> @syakesaba commented on GitHub (Aug 15, 2024): +1 Upvote for Japanese
Author
Owner

@gilacop commented on GitHub (Oct 13, 2024):

upvote for Hebrew

<!-- gh-comment-id:2408948693 --> @gilacop commented on GitHub (Oct 13, 2024): upvote for Hebrew
Author
Owner

@awilhelm-projects commented on GitHub (Oct 24, 2024):

I think you can close this - because we have https://ollama.com/library/bge-m3 and https://ollama.com/library/bge-large now. They are strong in multilanguage.

<!-- gh-comment-id:2433986065 --> @awilhelm-projects commented on GitHub (Oct 24, 2024): I think you can close this - because we have https://ollama.com/library/bge-m3 and https://ollama.com/library/bge-large now. They are strong in multilanguage.
Author
Owner

@larsschwarz commented on GitHub (Nov 15, 2024):

I think you can close this - because we have https://ollama.com/library/bge-m3 and https://ollama.com/library/bge-large now. They are strong in multilanguage.

@awilhelm-projects I've been trying these 2 on Linux and both exit with Error: "bge-large" does not support generate. No idea what's going on to be honest. Any idea what's the issue here?

<!-- gh-comment-id:2479432472 --> @larsschwarz commented on GitHub (Nov 15, 2024): > I think you can close this - because we have https://ollama.com/library/bge-m3 and https://ollama.com/library/bge-large now. They are strong in multilanguage. @awilhelm-projects I've been trying these 2 on Linux and both exit with `Error: "bge-large" does not support generate`. No idea what's going on to be honest. Any idea what's the issue here?
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#64261