I think there’s a balance to be struck here, step by step guides will teach you how to do one thing. But they deny you the experience of genuine problem solving that will enable you to dynamically break methods apart and use the parts that you need for a novel solution.
If you really lack the time, you can still use this technology through higher level methods such as the auto trainer.
If you want to dive lower, you’re gonna need to accept that this will take time. Took me ~3 months to learn the ins and outs of the Trainer and how it interacts with different model and data types. But I also had a fair bit of programming experience and exposure training loops at the pytorch level.
Getting good at this stuff is rewarding and valuable, but it’s not easy and it does require time investment.
The LLM cookbooks are good, so are the courses by huggingface. If you lack python experience then you need to learn python first (or accept that you will learn it on the go).