i would like to learn deeply about fine-tuning local models before burning money

There's so many new techniques like RL, RL LoRA, QLoRA, CPT LoRA.

I believe i would have a usecase for them, but i don't know where to learn, youtube is filled with bad quality tutorials if i just search and the good channels (fireship, bycloud) don't cover these as they're quite new concepts, i guess?.

how can a regular joe like me learn about these concepts in a "practical depth" so i can actually fine-tune qwen 27b successfuly on lets say custom corpus? without spending 100$ figuring out that "oh i didnt even need CPT here" or "well i chose the wrong Rank count! time to start this 2 day run again!"

context and TLDR: im building a legal general purpose chatbot for context, i have a big corpus, but im a bit stuck on what to do next

thanks for reading and any pointers!