Vicunia 2.0

[EliasVincent]

A lot has happened in the LLaMA / Alpaca / gpt4all world in the past two weeks. Many models have been released, each much better than the previous one.
With Vicunia 2.0, I want to have a good out-of-the-box experience for one of them. Here's the plan

• Vicunia comes with a pre-compiled .cpp binary for all platforms (with an option to bring your own)
• The "setup" will just be a button to download a model that works with the binaries
• Markdown rendering in the chat
• More fixes, of course

Right now, I think gpt4all might be the best to do this with. Let me know if a better one comes out in the meantime!