V2.0: Add more images for the training, and adjust the training caption to better match with other models.
This Lora training is based on the 768X1024 image, training on ProtogenInfinity checkpoint.
