Find answers from the community

Updated 3 months ago

llamaindex-embedding-lora/finetune_embed...

if we are using an llm for embeddings do we just pass the llm into the huggingface_https://github.com/marib00/llamaindex-embedding-lora/blob/main/finetune_embedding_lora.ipynb

9 comments

LLogan M

(why are you using an LLM for embeddings? 😅 )

It might work? I'm really not sure actually, I think it depends on what sentence-transformers is supporting

ddrewskidang

I'm just experimenting lol

ddrewskidang

and too lazy to do retriver agent memory, and I also want to add memory to embedding model for science lol..

LLogan M

Interesting idea!

ddrewskidang

this could be bro science if you ask like follow up questions like "explain more" the retriver models isn't gonna take into previous question rihgt

LLogan M

I would think so, if I understand LLM embeddings properly 👀

ddrewskidang

so what if its an LLM

LLogan M

LLMs are decoders, other embed models are encoders, so embeddings with LLMs are a special case

ddrewskidang

oh yeah i know lol.. i'm reffering to llms that have modied to do both... I think they are doing both lol https://huggingface.co/GritLM/GritLM-7B

Add a reply