Just installed FastChat with Vicuna weights [1].
This is basically a LLM with performance similar to GPT-4 (90% of GPT-4, according to the developers) [2].
The Vicuna weights are basically a "fine-tune" on top of LLaMA (Meta LLM) weights. They could achieve these performances doing the additional training on the ShareGPT dataset, which is a curated collection extracted from Chat-GPT real chats [3].
The beautiful thing is that it runs on a laptop: there is no remote API call involved. Everything runs on a 32GB medium/high-end laptop!
A bit tricky to install, but I am mind-blown.
β
[1] https://github.com/lm-sys/FastChat
[2] For now, I only used a limited version (7B), tomorrow will I try downloading 13B - https://vicuna.lmsys.org
[3] https://sharegpt.com
GitHub - lm-sys/FastChat: An open platform for training, serving, and evaluating large language models. Release repo for Vicuna and Chatbot Arena.
An open platform for training, serving, and evaluating large language models. Release repo for Vicuna and Chatbot Arena. - GitHub - lm-sys/FastChat: An open platform for training, serving, and eval...