Loss functions and Optimizer for the funetuning of GPT 3.5

I was wondering which are the loss function and the optimizer that are used in the finetuning of GPT 3.5, because I cannot find them in the documentation?