MOSS Image : Updated dataset

Anima version is trained with noise offset. Should handle dark scene/contrast a tad better.

Checkpik style lora. No trigger needed.