Helix v0.5 has been announced with improved text fine-tuning and OpenAI API support. The team faced challenges in fine-tuning the Mistral-7B model but successfully improved its performance. They chose fine-tuning over RAG for various reasons. Subscribe to HelixML for more updates.
Table of contents
tl;dr“The doctors are going to perform a tracheostomy”Document IDs aren’t just for RAGHow do you know when you’re good?Why not just use RAG?OpenAI API SupportOther small things:Sort: