Vicuna-13B: open-source chatbot trained by fine-tuning LLaMA on ChatGPT data

Delta weights have been released, can be automatically applied to the LLaMA weights: https://news.ycombinator.com/item?id=35430432