There’s a local llama subreddit with a lot of good information and 4chan’s /g/ board will usually have a good thread with a ton of helpful links in the first post. You can run some good models on a decent home but training and fine tuning will likely require renting out some cloud gpus.