Maeve (Paper Lyn Style) because I really like the cute art style
Recommend Prompt
Positive: maeve, tail, animal ears, source_anime,<lora:Maeve (Paper Lyn Style):1>
Negative: real life, photograph,
Description
Illustrious version
Comments (5)
It's really cute, but the LORA storage capacity is too large. Could you at least reduce it to about half?
you want me to quant the lora?
@Moonpaws Yes, I would like LORA to be made lighter.
@lone_warrior_tiv you can do that? wont that lose quality? image gen model is a LOT more suspectable to quant then llm
@lone_warrior_tiv you can do it yourself using the 'networks/resize_lora.py` script from kohya-ss' repository. And you won't lose much quality because already trained LoRAs compressed with this script retain much more 'knowledge' than LoRAs initially trained with low dim/alpha



