LoRA utente
Trained from Azusa's work, based on PonyXL v6.V02, cleaned up dataset, trained for more repeats.
