Build with >799a1cb13b0b1b560ab0ceff485caed68faa8f1f to support Mixtral

[sfxworks]

I believe, in order to resolve mudler/LocalAI#1446, go-llama.cpp needs to be built against at least version 799a1cb13b0b1b560ab0ceff485caed68faa8f1f of llama.cpp to enable mixtral support

[sfxworks]

I believe the auto prs are related
#313

[msuiche]

Any update on this

[allnash]

@mudler how can we help get this in? I get an error building on Apple Metal with the current master and this llama.cpp commit 799a1cb13b0b1b560ab0ceff485caed68faa8f1f

Any advice?