Thank you!
I was considering pinging you, but whatever you're running seems to snap up new releases :)
I had some quants made myself, but they failed for some reason.
Verified that at least the q3 works. So again, much obliged!
Whatever I'm running is my wetware going through the list of new models every day. And I overlook things, too, so it's fine to ping me when I seem to have skipped a model. Models also often don't work for whatever reason, and I usually quietly skip those, so pinging me will also mean you would find out what went wrong (although, of course, some people never make totally broken models :)
As for fails, it was quite the art to convert models beginning of the year (with lots of configurable options), but right now, llama is in a state where it either works or it doesn't, and there is pretty much nothing you can do to influence it. So if I can quantize it, so should you be able to - all you need is a more or less current llama.cpp, and (thats what trips most people), you need to install the python requirements (something like pip3 install -r requirements.txt, preferably in some venv, so some python knowledge is of course useful).
Anyway, good luck to you in the future :)