Large language models - Training locally and on server with knowledge transfer

I want to learn about LLM training, using existing models for training locally and on servers and the ability to provide API access to the model. I am familiar with oobabooga and was able to get it up and running locally with a ggml model.
Looking for somebody who is very good communicator and knowledgeable in LLM model training, running locally on windows, training with custom data. Also proficiency with text to sql conversion models is really helpful.