What about adding OpenLLaMA support?

[Sovenok-Hacker]

I think this is a very interesting model. The quality is a bit worse than the original LLaMA, but it can be used for commercial purposes. It took me about 2 hours to get it working with LLaMA.CPP. I used code from this gist to convert official files to GGML format and quantized it. It is a good and growing model. I think it would be good to add a guide on how to run it.

[Green-Sky]

You must have done something wrong, since it should just have worked.
see #1291

[Sovenok-Hacker]

Ok