Yasunori Ozaki
commited on
Commit
•
b1f91d8
1
Parent(s):
1417b95
Update README
Browse files- README.md +14 -15
- README_en.md +13 -14
README.md
CHANGED
@@ -14,7 +14,7 @@ tags:
|
|
14 |
English version is [here](README_en.md).
|
15 |
|
16 |
# はじめに
|
17 |
-
Cool Japan Diffusion
|
18 |
|
19 |
# ライセンスについて
|
20 |
ライセンスについては、もとのライセンス CreativeML Open RAIL++-M License に例外を除き商用利用禁止を追加しただけです。
|
@@ -38,9 +38,9 @@ Cool Japan Diffusion (for learning) はStable Diffsionをファインチュー
|
|
38 |
したがって、法的な側面ではなく、倫理的な側面を調査する目的も本配布は兼ねていると考えてください。
|
39 |
|
40 |
# 使い方
|
41 |
-
手軽に楽しみたい方は、こちらの[Space](https://huggingface.co/spaces/
|
42 |
-
詳しい本モデルの取り扱い方は[こちらの取扱説明書](https://alfredplpl.hatenablog.com/entry/
|
43 |
-
モデルは[ここ](https://huggingface.co/aipicasso/cool-japan-diffusion-2-1-
|
44 |
|
45 |
以下、一般的なモデルカードの日本語訳です。
|
46 |
|
@@ -71,7 +71,7 @@ Stable Diffusion v2と同じ使い方です。
|
|
71 |
|
72 |
### Web UIの場合
|
73 |
|
74 |
-
こちらの[取扱説明書](https://alfredplpl.hatenablog.com/entry/
|
75 |
|
76 |
### Diffusersの場合
|
77 |
|
@@ -86,20 +86,19 @@ pip install --upgrade git+https://github.com/huggingface/diffusers.git transform
|
|
86 |
次のスクリプトを実行し、画像を生成してください。
|
87 |
|
88 |
```python
|
89 |
-
from diffusers import StableDiffusionPipeline,
|
90 |
import torch
|
91 |
|
92 |
-
model_id = "aipicasso/cool-japan-diffusion-2-1-1
|
93 |
|
94 |
-
scheduler =
|
95 |
-
pipe = StableDiffusionPipeline.from_pretrained(model_id, scheduler=scheduler, torch_dtype=torch.float16)
|
96 |
pipe = pipe.to("cuda")
|
97 |
|
98 |
-
prompt = "anime, a portrait of a girl
|
99 |
-
negative_prompt="
|
100 |
-
|
101 |
-
|
102 |
-
image.save("girl.png")
|
103 |
|
104 |
```
|
105 |
|
@@ -174,7 +173,7 @@ Stable Diffusionと同じバイアスが掛かっています。
|
|
174 |
- VAEについて
|
175 |
- Danbooruなどの無断転載サイトを除いた日本の国内法を遵守したデータ: 60万種類 (データ拡張により無限枚作成)
|
176 |
- U-Netについて
|
177 |
-
- Danbooruなどの無断転載サイトを除いた日本の国内法を遵守したデータ:
|
178 |
|
179 |
**学習プロセス**
|
180 |
|
|
|
14 |
English version is [here](README_en.md).
|
15 |
|
16 |
# はじめに
|
17 |
+
Cool Japan Diffusion はStable Diffsionをファインチューニングして、アニメやマンガ、ゲームなどのクールジャパンを表現することに特化したモデルです。なお、内閣府のクールジャパン戦略とは特に関係はありません。
|
18 |
|
19 |
# ライセンスについて
|
20 |
ライセンスについては、もとのライセンス CreativeML Open RAIL++-M License に例外を除き商用利用禁止を追加しただけです。
|
|
|
38 |
したがって、法的な側面ではなく、倫理的な側面を調査する目的も本配布は兼ねていると考えてください。
|
39 |
|
40 |
# 使い方
|
41 |
+
手軽に楽しみたい方は、こちらの[Space](https://huggingface.co/spaces/aipicasso/cool-japan-diffusion-latest-demo)をお使いください。
|
42 |
+
詳しい本モデルの取り扱い方は[こちらの取扱説明書](https://alfredplpl.hatenablog.com/entry/2023/01/11/182146)にかかれています。
|
43 |
+
モデルは[ここ](https://huggingface.co/aipicasso/cool-japan-diffusion-2-1-1/resolve/main/v2-1-1.ckpt)からダウンロードできます。
|
44 |
|
45 |
以下、一般的なモデルカードの日本語訳です。
|
46 |
|
|
|
71 |
|
72 |
### Web UIの場合
|
73 |
|
74 |
+
こちらの[取扱説明書](https://alfredplpl.hatenablog.com/entry/2023/01/11/182146)に従って作成してください。
|
75 |
|
76 |
### Diffusersの場合
|
77 |
|
|
|
86 |
次のスクリプトを実行し、画像を生成してください。
|
87 |
|
88 |
```python
|
89 |
+
from diffusers import StableDiffusionPipeline, EulerAncestralDiscreteScheduler
|
90 |
import torch
|
91 |
|
92 |
+
model_id = "aipicasso/cool-japan-diffusion-2-1-1"
|
93 |
|
94 |
+
scheduler = EulerAncestralDiscreteScheduler.from_pretrained(model_id, subfolder="scheduler")
|
95 |
+
pipe = StableDiffusionPipeline.from_pretrained(model_id, scheduler=scheduler, torch_dtype=torch.float16)#,use_auth_token="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")
|
96 |
pipe = pipe.to("cuda")
|
97 |
|
98 |
+
prompt = "anime, masterpiece, a portrait of a girl, good pupil, 4k, detailed"
|
99 |
+
negative_prompt="deformed, blurry, bad anatomy, bad pupil, disfigured, poorly drawn face, mutation, mutated, extra limb, ugly, poorly drawn hands, bad hands, fused fingers, messy drawing, broken legs censor, low quality, mutated hands and fingers, long body, mutation, poorly drawn, bad eyes, ui, error, missing fingers, fused fingers, one hand with more than 5 fingers, one hand with less than 5 fingers, one hand with more than 5 digit, one hand with less than 5 digit, extra digit, fewer digits, fused digit, missing digit, bad digit, liquid digit, long body, uncoordinated body, unnatural body, lowres, jpeg artifacts, 3d, cg, text, japanese kanji"
|
100 |
+
images = pipe(prompt,negative_prompt=negative_prompt, num_inference_steps=20).images
|
101 |
+
images[0].save("girl.png")
|
|
|
102 |
|
103 |
```
|
104 |
|
|
|
173 |
- VAEについて
|
174 |
- Danbooruなどの無断転載サイトを除いた日本の国内法を遵守したデータ: 60万種類 (データ拡張により無限枚作成)
|
175 |
- U-Netについて
|
176 |
+
- Danbooruなどの無断転載サイトを除いた日本の国内法を遵守したデータ: 180万ペ���
|
177 |
|
178 |
**学習プロセス**
|
179 |
|
README_en.md
CHANGED
@@ -3,7 +3,7 @@
|
|
3 |
![eyecatch](eyecatch.jpg)
|
4 |
|
5 |
# Introduction
|
6 |
-
Cool Japan Diffusion
|
7 |
Cool Japan Diffusion is suitable for genetrating Cool Japan images such as Anime, Manga, and Game.
|
8 |
|
9 |
# Legal and ethical information
|
@@ -14,9 +14,9 @@ Therefore, we cannot use the model for commercially except for news reporting.
|
|
14 |
TBA.
|
15 |
|
16 |
# Usage
|
17 |
-
You can try the model by our [Space](https://huggingface.co/spaces/
|
18 |
I recommend to use the model by Web UI.
|
19 |
-
You can download the model [here](https://huggingface.co/aipicasso/cool-japan-diffusion-2-1-
|
20 |
|
21 |
## Model Details
|
22 |
- **Developed by:** Robin Rombach, Patrick Esser, Alfred Increment
|
@@ -42,7 +42,7 @@ You can download the model [here](https://huggingface.co/aipicasso/cool-japan-di
|
|
42 |
- Diffusers
|
43 |
|
44 |
## Web UI
|
45 |
-
Download the model [here]().
|
46 |
Then, install [Web UI](https://github.com/AUTOMATIC1111/stable-diffusion-webui) by AUTIMATIC1111.
|
47 |
|
48 |
## Diffusers
|
@@ -56,20 +56,19 @@ pip install --upgrade git+https://github.com/huggingface/diffusers.git transform
|
|
56 |
Running the pipeline (if you don't swap the scheduler it will run with the default DDIM, in this example we are swapping it to EulerDiscreteScheduler):
|
57 |
|
58 |
```python
|
59 |
-
from diffusers import StableDiffusionPipeline,
|
60 |
import torch
|
61 |
|
62 |
-
model_id = "aipicasso/cool-japan-diffusion-2-1-1
|
63 |
|
64 |
-
scheduler =
|
65 |
-
pipe = StableDiffusionPipeline.from_pretrained(model_id, scheduler=scheduler, torch_dtype=torch.float16)
|
66 |
pipe = pipe.to("cuda")
|
67 |
|
68 |
-
prompt = "anime, a portrait of a girl
|
69 |
-
negative_prompt="
|
70 |
-
|
71 |
-
|
72 |
-
image.save("girl.png")
|
73 |
|
74 |
```
|
75 |
|
@@ -78,4 +77,4 @@ image.save("girl.png")
|
|
78 |
- If you have low GPU RAM available, make sure to add a `pipe.enable_attention_slicing()` after sending it to `cuda` for less VRAM usage (to the cost of speed)
|
79 |
|
80 |
|
81 |
-
*This model card was written by: Alfred Increment and is based on the [Stable Diffusion v2](https://huggingface.co/stabilityai/stable-diffusion-2/raw/main/README.md)
|
|
|
3 |
![eyecatch](eyecatch.jpg)
|
4 |
|
5 |
# Introduction
|
6 |
+
Cool Japan Diffusion is the latent diffusion model created from Stable Diffsion.
|
7 |
Cool Japan Diffusion is suitable for genetrating Cool Japan images such as Anime, Manga, and Game.
|
8 |
|
9 |
# Legal and ethical information
|
|
|
14 |
TBA.
|
15 |
|
16 |
# Usage
|
17 |
+
You can try the model by our [Space](https://huggingface.co/spaces/aipicasso/cool-japan-diffusion-latest-demo).
|
18 |
I recommend to use the model by Web UI.
|
19 |
+
You can download the model [here](https://huggingface.co/aipicasso/cool-japan-diffusion-2-1-1/resolve/main/v2-1-1.ckpt). Safetensor version is [here](https://huggingface.co/aipicasso/cool-japan-diffusion-2-1-1/resolve/main/v2-1-1.safetensors). Japanese manual is [here](https://alfredplpl.hatenablog.com/entry/2023/01/11/182146).
|
20 |
|
21 |
## Model Details
|
22 |
- **Developed by:** Robin Rombach, Patrick Esser, Alfred Increment
|
|
|
42 |
- Diffusers
|
43 |
|
44 |
## Web UI
|
45 |
+
Download the model [here](https://huggingface.co/aipicasso/cool-japan-diffusion-2-1-1/resolve/main/v2-1-1.ckpt).
|
46 |
Then, install [Web UI](https://github.com/AUTOMATIC1111/stable-diffusion-webui) by AUTIMATIC1111.
|
47 |
|
48 |
## Diffusers
|
|
|
56 |
Running the pipeline (if you don't swap the scheduler it will run with the default DDIM, in this example we are swapping it to EulerDiscreteScheduler):
|
57 |
|
58 |
```python
|
59 |
+
from diffusers import StableDiffusionPipeline, EulerAncestralDiscreteScheduler
|
60 |
import torch
|
61 |
|
62 |
+
model_id = "aipicasso/cool-japan-diffusion-2-1-1"
|
63 |
|
64 |
+
scheduler = EulerAncestralDiscreteScheduler.from_pretrained(model_id, subfolder="scheduler")
|
65 |
+
pipe = StableDiffusionPipeline.from_pretrained(model_id, scheduler=scheduler, torch_dtype=torch.float16)#,use_auth_token="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")
|
66 |
pipe = pipe.to("cuda")
|
67 |
|
68 |
+
prompt = "anime, masterpiece, a portrait of a girl, good pupil, 4k, detailed"
|
69 |
+
negative_prompt="deformed, blurry, bad anatomy, bad pupil, disfigured, poorly drawn face, mutation, mutated, extra limb, ugly, poorly drawn hands, bad hands, fused fingers, messy drawing, broken legs censor, low quality, mutated hands and fingers, long body, mutation, poorly drawn, bad eyes, ui, error, missing fingers, fused fingers, one hand with more than 5 fingers, one hand with less than 5 fingers, one hand with more than 5 digit, one hand with less than 5 digit, extra digit, fewer digits, fused digit, missing digit, bad digit, liquid digit, long body, uncoordinated body, unnatural body, lowres, jpeg artifacts, 3d, cg, text, japanese kanji"
|
70 |
+
images = pipe(prompt,negative_prompt=negative_prompt, num_inference_steps=20).images
|
71 |
+
images[0].save("girl.png")
|
|
|
72 |
|
73 |
```
|
74 |
|
|
|
77 |
- If you have low GPU RAM available, make sure to add a `pipe.enable_attention_slicing()` after sending it to `cuda` for less VRAM usage (to the cost of speed)
|
78 |
|
79 |
|
80 |
+
*This model card was written by: Alfred Increment and is based on the [Stable Diffusion v2](https://huggingface.co/stabilityai/stable-diffusion-2/raw/main/README.md)
|