Update README.md
Browse files
README.md
CHANGED
|
@@ -28,7 +28,26 @@ It tries to come up with an appropriate composition using the specified content
|
|
| 28 |
The generated prompts are optimized for the image generation model [cagliostrolab/animagine-xl-4.0 opt](https://huggingface.co/cagliostrolab/animagine-xl-4.0)
|
| 29 |
生成されるプロンプトは画像生成モデル[cagliostrolab/animagine-xl-4.0 opt](https://huggingface.co/cagliostrolab/animagine-xl-4.0)に最適化してあります
|
| 30 |
|
| 31 |
-
###
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 32 |
|
| 33 |
I mainly train in Japanese, with English and Chinese (Traditional) as secondary languages.
|
| 34 |
I think it will probably work with other languages that are supported by qwen 2.5.
|
|
@@ -36,9 +55,9 @@ I think it will probably work with other languages that are supported by qwen 2.
|
|
| 36 |
日本語をメインにし、英語、中国語(繁体字)をサブでトレーニングをしています
|
| 37 |
おそらくqwen 2.5が対応している言語であれば他の言語でも動作するのではないかと思います。
|
| 38 |
|
| 39 |
-
|
| 40 |
|
| 41 |
-
|
| 42 |
|
| 43 |
It is necessary to provide it in Japanese as follows. The total input context length, including the prompt, should be around 6000.
|
| 44 |
以下のように日本語で与える必要があります。入力可能なコンテキスト長はプロンプトも含めてトータルで6000程度を目安としてください。
|
|
@@ -90,7 +109,7 @@ masterpiece, best quality, highresなどの品質に関連するタグは後工
|
|
| 90 |
- your tag list
|
| 91 |
```
|
| 92 |
|
| 93 |
-
|
| 94 |
|
| 95 |
thinkとpromptで囲まれて出力されます
|
| 96 |
think tag and prompt tag output.
|
|
@@ -102,14 +121,14 @@ think tag and prompt tag output.
|
|
| 102 |
詳細はSample Scriptを確認してください
|
| 103 |
Please check Sample Script for details.
|
| 104 |
|
| 105 |
-
|
| 106 |
以下のサンプルスクリプトを使うと、Googleが提供してくれているColabで無料で画像を生成する所まで確認する事ができます
|
| 107 |
Using the sample script below, you can check how to generate images for free using Colab provided by Google.
|
| 108 |
|
| 109 |
[Sample Colab Script](https://github.com/webbigdata-jp/python_sample/blob/main/FanFic_Illustrator_demo.ipynb)
|
| 110 |
|
| 111 |
|
| 112 |
-
|
| 113 |
We would like to thank the following people for providing related models/methods/tools.
|
| 114 |
以下の関連モデル/手法/ツールを提供してくださった皆さんに感謝します。
|
| 115 |
- Base Model: [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct)
|
|
@@ -117,7 +136,7 @@ We would like to thank the following people for providing related models/methods
|
|
| 117 |
- Training Methods: (DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models)[https://arxiv.org/abs/2402.03300]
|
| 118 |
- Traing script: [Unsloth](https://huggingface.co/unsloth).
|
| 119 |
|
| 120 |
-
|
| 121 |
|
| 122 |
- **Developed by:** dahara1@webbigdata
|
| 123 |
- **Model type:** text generation
|
|
|
|
| 28 |
The generated prompts are optimized for the image generation model [cagliostrolab/animagine-xl-4.0 opt](https://huggingface.co/cagliostrolab/animagine-xl-4.0)
|
| 29 |
生成されるプロンプトは画像生成モデル[cagliostrolab/animagine-xl-4.0 opt](https://huggingface.co/cagliostrolab/animagine-xl-4.0)に最適化してあります
|
| 30 |
|
| 31 |
+
### Comparison with other models/他のモデルとの比較
|
| 32 |
+
|
| 33 |
+
```
|
| 34 |
+
An illustration of the scene where Reimu is surprised by a giant crab that appears on the snow in the grounds of Hakurei Shrine.
|
| 35 |
+
```
|
| 36 |
+
|
| 37 |
+
|
| 38 |
+
|
| 39 |
+
Grok3
|
| 40 |
+
|
| 41 |
+
|
| 42 |
+
|
| 43 |
+
Gemini 2.0 Flash (Image Generation) Experimental
|
| 44 |
+
|
| 45 |
+
|
| 46 |
+
|
| 47 |
+
FanFic-Illustrator
|
| 48 |
+
|
| 49 |
+
|
| 50 |
+
## Supported languages/サポート言語
|
| 51 |
|
| 52 |
I mainly train in Japanese, with English and Chinese (Traditional) as secondary languages.
|
| 53 |
I think it will probably work with other languages that are supported by qwen 2.5.
|
|
|
|
| 55 |
日本語をメインにし、英語、中国語(繁体字)をサブでトレーニングをしています
|
| 56 |
おそらくqwen 2.5が対応している言語であれば他の言語でも動作するのではないかと思います。
|
| 57 |
|
| 58 |
+
## prompt template/プロンプトテンプレート
|
| 59 |
|
| 60 |
+
### Input/入力
|
| 61 |
|
| 62 |
It is necessary to provide it in Japanese as follows. The total input context length, including the prompt, should be around 6000.
|
| 63 |
以下のように日本語で与える必要があります。入力可能なコンテキスト長はプロンプトも含めてトータルで6000程度を目安としてください。
|
|
|
|
| 109 |
- your tag list
|
| 110 |
```
|
| 111 |
|
| 112 |
+
### Output/出力
|
| 113 |
|
| 114 |
thinkとpromptで囲まれて出力されます
|
| 115 |
think tag and prompt tag output.
|
|
|
|
| 121 |
詳細はSample Scriptを確認してください
|
| 122 |
Please check Sample Script for details.
|
| 123 |
|
| 124 |
+
## Sample Script
|
| 125 |
以下のサンプルスクリプトを使うと、Googleが提供してくれているColabで無料で画像を生成する所まで確認する事ができます
|
| 126 |
Using the sample script below, you can check how to generate images for free using Colab provided by Google.
|
| 127 |
|
| 128 |
[Sample Colab Script](https://github.com/webbigdata-jp/python_sample/blob/main/FanFic_Illustrator_demo.ipynb)
|
| 129 |
|
| 130 |
|
| 131 |
+
## Acknowledgements/謝辞
|
| 132 |
We would like to thank the following people for providing related models/methods/tools.
|
| 133 |
以下の関連モデル/手法/ツールを提供してくださった皆さんに感謝します。
|
| 134 |
- Base Model: [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct)
|
|
|
|
| 136 |
- Training Methods: (DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models)[https://arxiv.org/abs/2402.03300]
|
| 137 |
- Traing script: [Unsloth](https://huggingface.co/unsloth).
|
| 138 |
|
| 139 |
+
## Developer/開発
|
| 140 |
|
| 141 |
- **Developed by:** dahara1@webbigdata
|
| 142 |
- **Model type:** text generation
|