I would start with linear algebra, some calculus and statistics and understand how a neural network - which really is just one type of ML - works, the learn the basics of CNN and RNN, then learn transformers and LLM.

But that is just me. I think is more useful to understand the how and whys before training a LLM.