- Joined
- Apr 9, 2022
- Messages
- 891
- Reaction score
- 2,380
- Awards
- 11
Koboldcpp will run llama3 and is backward compatible to llama1. Version 1.66 is just out today. It is easy to install on Windows, Linux, and MacOS. I simply download the Linux executable, put it in my /home directory, give it permission to run (in user space), and launch it via a .desktop file I made. It's likely similar in Windows. You just download the single .exe file, put it somewhere convenient, and click to run it. You can even compile the program yourself because it is open-source software. Aside from versions which enable NVidia CUDA acceleration, there is a branch for AMD (ROCm), Vulcan compute, and CPU-only modes of operation. Find tons of GGUF models on Hugging Face.
And a very active Discord community:
(Discord link will expire within 7 days of this initial post.)
And a very active Discord community:
(Discord link will expire within 7 days of this initial post.)