Are there communities where one can go to learn more about fine tuning and running these things? I've found a bunch for diffusion models but haven't had any luck with LLMs.
The discord servers for a few of the projects are relatively popular. Most have a help channel you could post in if you have questions. The Discord for KoboldAI has some developers from koboldcpp ,which is the easiest and one of the most bleeding edge way of running these models locally. It builds on llamacpp and allows the use of different front ends among other things like using k quantized models. People also have had success with using something like runpods.
Native fine tuning is still out of consumer reach for the forseeable future, but there's people experimenting with QLORAs. The pipeline is still relatively new though and is a bit involved.