* build the vocab from the local dataset, split into the train, test, and validation * load the pre-trained model from HuggingFace * supervised fine-tuning with pre-trained models Referenced base models from #2
