Hey, tried to train my Girls und Panzer dataset consisting of Megami Magazine artworks of the franchise on Moss.
Works the best together with Lodestone's 8step dmd lora:
https://huggingface.co/lodestones/MOSS-Image-SFT-8-steps-dmd/blob/main/lora_latest.safetensors
Recommended settings:
CFG: 2 (yes, not 1. You really want the negative prompt to work for this model)
Steps: 8
Euler + Simple or whatever makes sense
Triggerword: Girls und Panzer
or Girls und Panzer anime illustration
LoRa Strength: 0.8
Description
Trained on MOSS Image Base and tested on Base, not sft.