Skip to content

A Chat GUI for local Alpaca language model with guided download and install instructions

License

MIT, MIT licenses found

Licenses found

MIT
LICENSE
MIT
LICENSE-gpt4all.md
Notifications You must be signed in to change notification settings

EliasVincent/vicunia

Repository files navigation

Vicunia - Alpaca.cpp Chat Frontend and Tooling

Vicunia is a frontend for using alpaca.cpp and providing a GUI for installing and chatting with Stanford Alpaca and other models from the llama.cpp family. It's available for Windows, Linux and Mac

Vicunia

Download

See releases for the latest version.

Setup Automatic (v2.0+)

  • Go to the "Setup" tab
  • Click "Download Model"
  • In the options menu in the chat tab, make sure the path points to the correct folder (vicunia root dir + resources/models)

On Windows, the vicunia root dir looks something like this: C:\Users\yourusername\AppData\Local\Programs\vicunia

Setup Manual

Follow the guide on alpaca.cpp!

TL;DR: have a working and compiled alpaca.cpp folder with a model in your OS home directory (Windows: C:\Users\%USERNAME%\alpaca.cpp, Linux: /home/%USERNAME%/alpaca.cpp, macOS: /Users/%USERNAME%/alpaca.cpp). The compiled binaries (chat.exe, etc..) and model should be in the root of the alpaca.cpp folder.

Troubleshooting

  • If current Alpaca.cpp throws errors, I have tested branches on my fork that work with Vicunia.
  • On Windows, CPU with AVX2 support is required (at least with the included binaries).
  • If you update your version, try deleting .vicunia-settings.json from the home folder first to avoid issues.
  • Mac version is not really tested right now

Feel free to open an issue on Github!

TODO

  • more options
  • markdown renderer

How to use other .cpp forks

  • rename executable to chat (OSX / Linux) or chat.exe
  • have model file in same folder
  • paste path to chat folder in the options menu

Credits

vicunia-logo