Sign In

Yamada Elf | Eromanga-sensei | SDXL

91
717
1.2k
11
Verified:
SafeTensor
Type
LoRA
Stats
537
1,025
29
Reviews
Published
Jan 29, 2024
Base Model
Pony
Training
Steps: 3,475
Epochs: 5
Usage Tips
Clip Skip: 1
Trigger Words
yamada_elf
pink bikini, striped bikini, side-tie_bikini_bottom, red hairbow, blonde hair, long hair, drill hair, ponytail hair, brown eyes, pointy ears
pink dress, bow, red hairbow, hairband, blonde hair, long hair, drill hair, brown eyes, pointy ears
Hash
AutoV2
7004FBAF6F

Knows two types of clothes: dress and bikini

For PonyDiffusion V6 be sure to use the tag score_9

The strength of LoRA depends on the prompt. Typically, the larger and more complex it is, the more force required, up to 1.0. But for simple and short prompts it is better to use a strength of 0.7.

Model: animagine-xl-3.0.safetensors or ponyDiffusionV6XL_v6.safetensors

Dataset size: 109

Clip skip: 2 (no difference with clip1)

dim32, alpha16

lr=0.0003, constant, adafactor

Optimizer extra arguments: scale_parameter=False relative_step=False warmup_init=False

version for SD1.5\NAI click

My tips for generating on SDXL for AUTOMATIC1111

Resolution:

  • 1024 x 1024

  • 1152 x 896 or 896 x 1152

  • 1216 x 832 or 832 x 1216

Don't use Hires. fix. Send the picture to img2img, set CFG 0.15-0.25, then select “SD upscale” in the script, you also need to select 4x-AnimeSharp or what you like best from upscale. All other settings are default. Click generate.

For me 12gb VRAM is enough for this using --xformers