<h1 id="heading-1817">Intro</h1>This clothing LoRA generate gradient or two-tone pantyhose. Works on both anime and realistic checkpoints. 这是一个 SD1.5 服饰LoRA，用于生成彩色的渐变丝。二次元/写实大模型均可使用，查看示例图片所使用的模型。Main trigger word 主要触发词 :two-tone legwear,then pick from The available colors 再搭配颜色触发词 :💚 green_pantyhose,💛 yellow_pantyhose,💙 blue_pantyhose,💜 purple_pantyhose,🩷 pink_pantyhose,<s>🖤 black_pantyhose，</s> doesn't quite work（很难生效）You can also try other colors and see if it works out well. 即可产生效果，可自行尝试其他颜色。<h1 id="heading-1818">Settings</h1>1. LoRA Block Weight SettingBy now I assume that everyone knows what LBW (LoRA Block Weight) is so I'll just put my preset here. 我将假设你已经知道如何使用LBW插件（在线生成服务不支持）从而防止LoRA对于画风、姿态或背景带来的污染，因此我会直接给出LBW预设。<a target="_blank" rel="ugc" href="https://github.com/hako-mikan/sd-webui-lora-block-weight">LBW文档</a>Preset 预设 (copy to your LBW weight setting box):StocTest:0.9,0,0.4,0,0,0,0.4,0,0.5,1,0.8,0,0,0,0,0,0How to use 使用方法 :The prompt has changed (as in webUI version &gt; 1.5). Now you must add ":lbw=XLFace" right after the weight of LoRA.e.g. &lt;lora:gradient_v1.0:1:1:lbw=StocTest&gt;将以上预设粘贴至LoRA Block Weight插件的权重设置中，保存并重载预设和标签；依照LoRA名称，填写&lt;lora:gradient_v1.0:1:1:lbw=StocTest&gt;请关注上述LoRA引用格式中的划线部分，1:1取代了原本填写LoRA权重的单个数值，这个比值的意思是Text-Encoder与U-Net的比，例如1:1 意味着“BASE权重为1.0，IN00 - OUT11权重均为1.0”，同理，1:0.5意味着“BASE权重为1.0，IN00 - OUT11权重均为0.5”。Regarding the weird "1:1" parameter, see <a target="_blank" rel="ugc" href="https://github.com/hako-mikan/sd-webui-lora-block-weight/issues/126">issue#126</a>2. General Settings 其他设置Step: 36Sampler: DPM++ 2M SDE KarrasCFG Scale: 7Resolution: 512*768don't use hi-res fix or refiner, not necessary. hi-res fix 或 refiner 均不需要使用。<h1 id="heading-1818">Training</h1><pre><code>Images: 1024*1536 px, 160 pics,
Tagging: wd-14 auto tagging with manual correction,
min_snr_gamma: 5,
offset_noise: 0.1,
network_dim: 32,
network_alpha: 16,</code></pre>

Files with SHA256: 67953b40da11f264f0d2504198657043c934a85c1bc09c0a794b465880d1567c