aparatext

> I'm interested in this too. Meanwhile I'm using Facebook StarSpace for sentence embeddings. It's very fast and easy to use. The official version, however, has some issues which are...

> > That's cool! But does LLaMA have a tiny version similar to OpenAI Ada to avoid wasting resources? I don't think most use-cases need anything more then BERT, which...

> Is there any other way to get the weights? I guess there's OpenLLaMA, which aims to retrain the weights from scratch...

The best thing to speed up ingestion would be to abandon the idea of using LLaMA for embeddings. Just like using full GPT-3 davinci to generate embeddings is costlier and...

> > The best thing to speed up ingestion would be to abandon the idea of using LLaMA for embeddings. Just like using full GPT-3 davinci to generate embeddings is...

> Cool concept but I think ATM using (local) vector stores and GPT 3 or 3.5 may be more practical (I know, not air gapped- whatever, write a proxy to...

I'm currently working on a similar project to realize at least the basics of this vision. For now, I'm getting the UI, DB, and soon embeddings + LLaMA to work....

Chances are, it's already partially using the GPU. As it is now, it's a script linking together LLaMa.cpp emeddings, Chroma vector DB, and GPT4All. GPT4All might be using PyTorch with...

Just replace the embedding model with [CodeBERT](https://github.com/microsoft/CodeBERT).

It seems there are other, less niche models for spelling correction, like [t5-spellchecker](https://huggingface.co/Bhuvana/t5-base-spellchecker) or [other BERT-based models](https://github.com/neuspell/neuspell). Since there's been [some work on T5](#12) and there [BERT.cpp](https://github.com/skeskinen/bert.cpp) (which does not...