/upscale
score_9, source_anime BREAK, oldest, anime screencap, outdoors, city, city lights, night, night sky, <lora:RoseAnderson(Pony)-11:1>RoseAnderson, 1girl, solo, long hair, blonde hair, ahoge, huge ahoge, capelet, collared capelet, shirt, blue shirt, buttons, puffy sleeves, long sleeves, (bridal gauntlets), boots, white boots, skirt, multicolored skirt, handgun, holding handgun, weapon, firing, sitting, spread legs, ((nsfw, panties aside, pussy)), light smile,
Vicioso69
×
One Trainer:
30 imagenes - screenshots
Profile default settings= SDXL 1.0 LoRA
BASE: Stable Diffusion XL Base
Type= LoRA (top right corner in one training)
In concept settings=
balancing [ 10.0 ] Repeats
On training settings:
Learning Rate Scheduler [Cosine with Restard],
Bath Size [1]->bfloat16
Epoch/Etapas: [12]->bfloat16,
On model settings:
Weight data type: bfloat16
Output data type: bfloat16
El resto de los parámetros los dejas con la configuración predeterminada del perfil SDXL-LoRA
You leave the rest of the parameters with the default configuration of the SDXL-LoRA profile