Update README.md
Browse files
README.md
CHANGED
@@ -1,5 +1,107 @@
|
|
1 |
-
---
|
2 |
-
license: other
|
3 |
-
license_name: faipl-1.0-sd
|
4 |
-
license_link: https://freedevproject.org/faipl-1.0-sd/
|
5 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: other
|
3 |
+
license_name: faipl-1.0-sd
|
4 |
+
license_link: https://freedevproject.org/faipl-1.0-sd/
|
5 |
+
language:
|
6 |
+
- en
|
7 |
+
tags:
|
8 |
+
- stable-diffusion
|
9 |
+
- sdxl
|
10 |
+
- anime
|
11 |
+
---
|
12 |
+
|
13 |
+
# Hello everyone and John6666, DO NOT RE-UPLOAD THIS MODEL
|
14 |
+
|
15 |
+
![image/jpeg](https://cdn-uploads.huggingface.co/production/uploads/630e2d981ef92d4e37a1694e/0ajhwTcu8uIcXCWvmeM6o.jpeg)
|
16 |
+
|
17 |
+
# Hikari Noob v-pred 0.6.5
|
18 |
+
|
19 |
+
Civitai model page: https://civitai.com/models/938672
|
20 |
+
|
21 |
+
Fine-tuned NoobAI-XL(ν-prediction) and merged SPO LoRA
|
22 |
+
|
23 |
+
NoobAI-XL(ν-prediction)をファインチューンし、SPOをマージしました。
|
24 |
+
|
25 |
+
## Features/特徴
|
26 |
+
- Improved stability and quality.
|
27 |
+
- Good results with only 10 steps (12 steps or more recommended)
|
28 |
+
- Fixed a problem in which the quality of output was significantly degraded when the number of tokens exceeded 76.
|
29 |
+
- The base style is not strong and can be restyled by prompts or LoRAs.
|
30 |
+
- This model does not include any base model other than NoobAI (ν-prediction 0.5), so it has the equivalent knowledge.
|
31 |
+
You can generate characters that have appeared by August 2024.
|
32 |
+
- 安定性と品質を改善
|
33 |
+
- わずか10ステップでよい結果を得られます(ただし12ステップ以上を推奨)
|
34 |
+
- トークン数が76を超えると出力の品質が著しく低下する問題を修正しました。
|
35 |
+
- 素の画風は強くないので、プロンプトやLoRAによる画風変更ができます。
|
36 |
+
- このモデルはNoobAI(ν-prediction 0.5)以外のベースモデルを一切含まず、それと同等の知識があります。2024年8月までに登場したキャラクターを生成できます。
|
37 |
+
|
38 |
+
## About v0.6.5
|
39 |
+
- Based on NoobAI-XL v-prediction 0.6
|
40 |
+
- Enabled Zero Terminal SNR
|
41 |
+
|
42 |
+
## Requirements / 動作要件
|
43 |
+
- AUTOMATIC1111 WebUI on `dev` branch / devブランチ上のAUTOMATIC1111 WebUI
|
44 |
+
- **Latest version** of ComfyUI / **最新版**のComfyUI
|
45 |
+
- **Latest version** of Forge or reForge / **最新版**のForgeまたはreForge
|
46 |
+
|
47 |
+
### Instruction for AUTOMATIC1111 / AUTOMATIC1111の導入手順
|
48 |
+
1. Switch branch to `dev` (Run this command in the root directory of the webui: `git checkout -b dev origin/dev` or use Github Desktop)
|
49 |
+
2. Use the model as usual!
|
50 |
+
|
51 |
+
(日本語)
|
52 |
+
1. `dev`ブランチに切り替えます(次のコマンドをwebui直下で実行します: `git checkout -b dev origin/dev` またはGithub Desktopを使う)
|
53 |
+
2. 通常通りモデルを使用します。
|
54 |
+
|
55 |
+
## Prompt Guidelines / プロンプト記法
|
56 |
+
Almost same as the base model/ベースモデルとおおむね同じ
|
57 |
+
|
58 |
+
To improve the quality of background, add `simple background, transparent background` to Negative Prompt.
|
59 |
+
|
60 |
+
## Recommended Prompt / 推奨プロンプト
|
61 |
+
Positive: None/無し(Works good without `masterpiece, best quality` / `masterpiece, best quality`無しでおk)
|
62 |
+
|
63 |
+
Negative: `worst quality, low quality, bad quality, lowres, jpeg artifacts, unfinished, photoshop \(medium\), abstract`, Leaving negative empty will break the output! Use `_` instead.
|
64 |
+
|
65 |
+
|
66 |
+
## Recommended Settings / 推奨設定
|
67 |
+
Steps: 8-24
|
68 |
+
|
69 |
+
Sampler: Euler(DPM++ 2M will work but unstable, other samplers will not work properly)
|
70 |
+
|
71 |
+
Scheduler: Simple
|
72 |
+
|
73 |
+
Guidance Scale: 2-7
|
74 |
+
|
75 |
+
### Hires.fix
|
76 |
+
|
77 |
+
Hires upscaler: 4x-UltraSharp or Latent(nearest-exact)
|
78 |
+
|
79 |
+
Denoising strength: 0.4-0.5(0.65-0.7 for latent)
|
80 |
+
|
81 |
+
|
82 |
+
## Merge recipe(Weighted sum)
|
83 |
+
- Stage 1: Finetune Hikari Noob v-pred 0.5 and merge(see below)
|
84 |
+
*A-F: noobai(v-pred)-based custom checkpoint
|
85 |
+
- A * 0.6 + B * 0.4 = tmp1
|
86 |
+
- tmp1 * 0.6 + C * 0.4 = tmp2
|
87 |
+
- tmp2 * 0.7 + F * 0.3 = tmp3
|
88 |
+
- tmp3 * 0.7 + E * 0.3 = tmp4
|
89 |
+
- tmp4 * 0.5 + D * 0.5 = tmp5
|
90 |
+
- tmp5 * 0.65 + G * 0.35 = tmp6
|
91 |
+
- Make H,I from tmp6
|
92 |
+
- tmp6 * 0.75 + H * 0.25 = tmp7
|
93 |
+
- tmp7 * 0.75 + I * 0.25 = tmp8
|
94 |
+
- tmp8 + SPO LoRA * 1 + sdxl-flat * -0.25 + sdxl-boldline * -1 = tmp9
|
95 |
+
- Adjust tmp9(0.2,0.2,0.2,0.1,0,0,0,0) = Result
|
96 |
+
|
97 |
+
## Training scripts:
|
98 |
+
[sd-scripts](https://github.com/kohya-ss/sd-scripts)
|
99 |
+
|
100 |
+
## Notice
|
101 |
+
This model is licensed under [Fair AI Public License 1.0-SD](https://freedevproject.org/faipl-1.0-sd/)
|
102 |
+
|
103 |
+
If you make modify this model, you must share both your changes and the original license.
|
104 |
+
|
105 |
+
You are prohibited from monetizing any close-sourced fine-tuned / merged model, which disallows the public from accessing the model's source code / weights and its usages.
|
106 |
+
|
107 |
+
Do not reupload this model.
|