Darkest Dungeon - [STYLE] [LORA] [FLUX] - V2.0 [26.11.2024]

Much less overtraining and smaller size by changing the training config by going from 32/32 dim/alpha to 16/16, removing caption dropout, and adding a 1e-7 Clip L text encoder, reducing training image count from 30 to 15, and generating new captions by ChatGPT and editing them