Local RAG - which embedding to use?

[ianscrivener]

I'm coding a RAG demo with llama.cpp, Weaviate vector database and LlamaIndex.. (which works closely with langchain).

Should I use llama.cpp embeddings, or a leading embedding model like BAAI/bge-small-en?

[Alumniminium]

i use bge-large-en-1.5 and it works very well.

[ianscrivener]

thanks @Alumniminium

[moatftw]

May I ask how do you use bge-large-en-1.5 in llama.cpp? Since I can't find it in the GGUF format that can be loaded in llama.cpp's embedding example.

[ianscrivener]

@Alumniminium ?

[Alumniminium]

Im not using it in llama.cpp im using it together with llama.cpp.