Downloads · 30 days
0
AinzOoalGown
samjwu/AinzOoalGown
AinzOoalGown is a machine learning model from samjwu. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as gpl-3.0.
Low Rank Adaptation for Ainz Ooal Gown from the Overlord series.
Access
Public
Updated Apr 15, 2023
Repo size
37.9 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors37.9 MB · 85%
At a glance
- License
- gpl-3.0
- Access
- Public
- Created
- Apr 5, 2023
- Updated
- Apr 15, 2023
- SHA
- ef3bd858
Model card
From the Hugging Face model README
Ainz Ooal Gown LoRA
Description
Low Rank Adaptation for Ainz Ooal Gown from the Overlord series.
A reference image from the light novel. See other references for the anime/manga in the references folder
Usage
If using Stable Diffusion WebUI:
- Download
AinzOoalGownLoRA.safetensors - Copy and paste into
stable-diffusion-webui/models/Lora - Use the trigger word
<lora:AinzOoalGown:0.8>in the prompt. Works best with weights ranging from 0.5 to 0.8 (default is 1)
Examples
The Good
Loses some likeness of the original character, but retains that cool skeletal undead look. See more (good) examples in the examples folder
Parameters:
Prompt: 1boy, (overlord, lich, skeleton, skull, bone, ainz ooal gown:1.2), (best quality:1.2), (masterpiece:1.2), (highres:1.2), extremely detailed, illustration, solo, (looking at viewer:1.2), (sharp focus:1.5), (masterpiece illustration:1.5), light particles, glowing light, lens flare, <lora:AinzOoalGown:0.8>
Negative prompt: EasyNegative, worst quality, low quality, normal quality, blurry, deformed, disfigured, morbid, mutated, bad anatomy, bad art, lowres, polar lowres, bad face, bad hands, bad body, bad feet, bad proportions, bad leg, more legs, gross proportions, poorly drawn, (text:1.5), error, missing fingers, missing arms, missing legs, short legs, extra digit, 2girls, 1girl, long skirt, indoors, out of frame, only upper body, only lower body, petite, print, chibi, only head
Steps: 30, Sampler: DDIM, CFG scale: 7, Seed: 4149917258, Size: 512x512, Model hash: 89d59c3dde, Model: nai
The Bad
Still interesting, but misshapen
The Ugly
It kinda looks like him if you squint
Training Details
- Training Model: Anything V3.1
- Training Dataset: 13 images
- Training Repeats: 10
- LoRA Network Linear Dimension: 32
- LoRA Network Linear Alpha: 16
- Optimizer: AdamW8bit
- UNet Learning Rate: 1e-4
- Text Encoder Learning Rate: 5e-5
- Learning rate warmup steps: 0
- Learning rate Scheduler: constant
- Sampler: ddim
- Epochs: 10
- Training Batch Size: 6
- Precision: fp16 Mixed
- Clip Skip: 2
- Gradient Accumulation Steps: 1
- Total Steps: 230
- License
- gpl-3.0
- Created
- Apr 5, 2023
- Updated
- Apr 15, 2023