Resumed from <a target="_blank" rel="ugc" href="https://huggingface.co/OnomaAIResearch/Illustrious-xl-early-release-v0">Illustrious XL</a> v0.1 by makkishizu.Mainly trained on high-rated Danbooru 2024 images, using 40k images on a 4060Ti over 2+8 epochs.Due to computing limitations, I did not train on an extremely large dataset. The primary goal was to improve the “default art style”(no artists tag) of Illustrious XL while tracking user preferences. Refer to the following comparison sample images:<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/be7d5567-7409-47a8-a372-128c3c89ff97/width=525/be7d5567-7409-47a8-a372-128c3c89ff97.jpeg" />If you prefer models trained on larger datasets, you can try <a target="_blank" rel="ugc" href="https://civarchive.com/models/833294/noobai-xl-nai-xl">noobai-xl-nai-xl</a>.<h2 id="usage-guidelines-b7irw52mm">Usage Guidelines</h2>This model is trained using Danbooru-style tags, not natural language! Please use tags only for optimal results.The usage is basically the same as Illustrious XL, you can adjust the art style using the artist list. If you are interested in the artist list but don't know how to write it, you can try using <a target="_blank" rel="ugc" href="https://github.com/MakkiShizu/stable-diffusion-webui-wildcards/blob/master/wildcards/novelai_300artist.txt">novelai_300artist</a>This model supports resolutions from ARB 1024x1024, with a minimum resolution of 256 and a maximum resolution of 2048. While standard SDXL resolution can be used, it is recommended to opt for a slightly higher resolution than 1024x1024. Applying a hires-fix is also suggested for better output quality.For more details, check the sample images provided.In addition to following the usage instructions of Illustrious XL:<pre><code>Recommended sampling method: Euler a, Sampling Steps: 20–28, CFG: 5–7.5 (may vary based on use case).
The model supports quality tags such as: "worst quality," "bad quality," "average quality," "good quality," "best quality," and "masterpiece (quality)."</code></pre>This model also balances the distribution of rating tags, allowing you to distinguish images by different rating levels:<pre><code>Rating Modifier	 Rating Criterion
safe	 General
sensitive	 Sensitive
nsfw	 Questionable
explicit, nsfw	 Explicit</code></pre>Recommended prompt format:<pre><code>&lt;|special|&gt;, 
&lt;|characters|&gt;, &lt;|copyrights|&gt;, 
&lt;|artist|&gt;, 

&lt;|general|&gt;, 

&lt;|quality|&gt;, &lt;|meta|&gt;, &lt;|rating|&gt;</code></pre>Recommended Negative Prompt:<pre><code>worst quality, comic, multiple views, bad quality, low quality, lowres, displeasing, very displeasing, bad anatomy, bad hands, scan artifacts, monochrome, greyscale, twitter username, jpeg artifacts, 2koma, 4koma, guro, extra digits, fewer digits, jaggy lines, unclear</code></pre><h2 id="training-details:-09ft79pkn">Training Details:</h2>The dataset for training this model was sourced from <a target="_blank" rel="ugc" href="https://github.com/KohakuBlueleaf/HakuBooru.git">hakubooru</a>.The training of MakkiXL was facilitated by the LyCORIS project and the trainer from lora-scripts.The original LoKr file is also provided as the "makki_illustrious_lokr" version. For detailed settings, refer to the LyCORIS config file from makki_illustrious_lokr.<pre><code>Hardware: RTX 4060Ti

Num Train Images: 43,216

Total Epoch: 2+8

Total Steps: 6760

Batch Size: 1

Grad Accumulation Step: 64

Equivalent Batch Size: 64

Optimizer: Lion8bit

Learning Rate: 5e-5 for UNet /NO train TE

LR Scheduler: Constant

Min SNR Gamma: 5

Noise Offset: 0.03

Resolution: 1024x1024

Min Bucket Resolution: 256

Max Bucket Resolution: 2048

Mixed Precision: BF16</code></pre><h2 id="license-kzear8y2d">License</h2>This model is released under the Fair-AI-Public-License-1.0-SD.Please check this website for more information:Freedom of Development <a target="_blank" rel="ugc" href="http://freedevproject.org">freedevproject.org</a><h2 id="contributors'-repositories-7w6oc4njg">Contributors' Repositories</h2><a target="_blank" rel="ugc" href="https://github.com/Akegarasu/lora-scripts.git">lora-scripts</a><a target="_blank" rel="ugc" href="https://github.com/KohakuBlueleaf/HakuBooru.git">hakubooru</a>Thanks to onommai open source for providing such a powerful base model.

MakkiXL

v0.1: 中文版本MakkiXL基于 <a target="_blank" rel="ugc" href="https://huggingface.co/OnomaAIResearch/Illustrious-xl-early-release-v0">Illustrious XL</a> v0.1版本进行训练。主要针对 Danbooru 2024 的高评分图片进行训练，使用 4w 张图片在 4060Ti 上训练了 2+8 个 epoch。由于算力限制，我并没有使用特别大量的数据集进行训练。最初目的为以相对用户偏好改善 Illustrious XL 的“默认画风”（无画师串）。如果你喜欢使用更大量数据集训练的模型，可以尝试使用 <a target="_blank" rel="ugc" href="https://civitai.com/models/833294/noobai-xl-nai-xl">noobai-xl-nai-xl</a>。使用指南该模型使用的是 Danbooru 的标签形式进行训练，而非自然语言！请务必以 tag only 的形式进行使用。用法与 Illustrious XL 基本相同，你可以使用画师列表对画风进行调整。（如果你对画师列表感兴趣但不知道如何编写，可以尝试使用<a target="_blank" rel="ugc" href="https://github.com/MakkiShizu/stable-diffusion-webui-wildcards/blob/master/wildcards/novelai_300artist.txt">novelai_300artist</a>）该模型以 1024x1024 分辨率为基准，最低分辨率为 256，最高分辨率为 2048。虽然可以使用标准 SDXL 分辨率，但建议使用比 1024x1024 稍高的分辨率。同时建议应用 hires-fix 以获得更好的结果。如需更多详细信息，请查看提供的示例图片。除了遵循 Illustrious XL 公开的使用方法外：<pre><code>推荐的采样方法：Euler a，采样步数：20–28，CFG：5–7.5（可能因使用场景不同而变化）。
该模型支持质量标签，如："worst quality," "bad quality," "average quality," "good quality," "best quality," 和 "masterpiece (quality),"</code></pre>该模型还针对评分的分布进行了平衡，您可以使用以下评分标签来区分不同评分的图片：<pre><code>评分修改	 评分标准
safe	 安全
sensitive	 敏感
nsfw	 色色
explicit, nsfw	 色色+</code></pre>推荐的提示词格式：<pre><code>&lt;|special|&gt;, 
&lt;|characters|&gt;, &lt;|copyrights|&gt;, 
&lt;|artist|&gt;, 

&lt;|general|&gt;, 

&lt;|quality|&gt;, &lt;|meta|&gt;, &lt;|rating|&gt;</code></pre>推荐的负面提示词:<pre><code>worst quality, comic, multiple views, bad quality, low quality, lowres, displeasing, very displeasing, bad anatomy, bad hands, scan artifacts, monochrome, greyscale, twitter username, jpeg artifacts, 2koma, 4koma, guro, extra digits, fewer digits, jaggy lines, unclear</code></pre>训练内容详情：该模型的训练数据集来源于 <a target="_blank" rel="ugc" href="https://github.com/KohakuBlueleaf/HakuBooru.git">hakubooru</a> 。MakkiXL 使用LoKr形式进行训练，我将同时提供原始的 LoKr 文件，名为 "makki_illustrious_lokr"。有关详细设置，请参考 makki_illustrious_lokr 的元数据。<pre><code>硬件：RTX 4060Ti

训练图片数量：43216

总 Epoch：2+8

总步数：6760

批量大小：1

梯度累积步数：64

等效批量大小：64

优化器：Lion8bit

学习率：5e-5 用于 UNet / 不训练 TE

学习率调度器：Constant

最小 SNR Gamma：5

噪声偏移：0.03

分辨率：1024x1024

最小 分桶 分辨率：256

最大 分桶 分辨率：2048

混合精度：BF16</code></pre>许可证该模型在 Fair-AI-Public-License-1.0-SD 许可证下发布。请访问以下网站了解更多信息：<a target="_blank" rel="ugc" href="http://freedevproject.org">freedevproject.org</a>贡献者的仓库<a target="_blank" rel="ugc" href="https://github.com/Akegarasu/lora-scripts.git">lora-scripts</a><a target="_blank" rel="ugc" href="https://github.com/KohakuBlueleaf/HakuBooru.git">hakubooru</a>感谢 onommai 开源提供了如此强大的基础模型。

Files with SHA256: 1222611f04821e37550143d80301ab113b27e33527cee33f667e429f8648c16a