no but i have thought about doing LoRas, my hardware is powerful enough for that in general i think it's better to enhance base models with your datasets --- but to answer your question, you need some python/ML knowledge, a dataset to train on, and as much compute as you can throw at training