<h2 id="heading-668">foreword（基本介绍）</h2>It is a Chinese ink style Lora model<ul><li>Best practice: arbitrary resolution, better Lora weight 0.5-1, clip=2, sampler DMP ++2M karras, 20-40step, using Chinese ink painting to call lora, more parameters can refer to my sketch</li></ul>这是一个中国水墨风格的lora模型<ul><li>最佳实践：分辨率随意，Lora权重0.5-1更好，clip=2，采样器DMP ++2M karras，20-40step，使用Chinese ink painting调用lora，更多参数可以参考我的示意图</li></ul><h2 id="heading-669">CONCEPT（概念）</h2>可以搭配任意大模型使用，这个版本更适合远景，全身，当然看你喜好，目前测试如果你想只想要一点点的效果但想要有影响物体我建议Lora值是0.2-0.4这样也许会有一点效果，但不至于太强烈，目前还处于基础版本，后续可能会更多的迭代与更新Lora我是基于anything大模型训练的lora，因此我认为他的适配强度很高目前我在测试Lora适配模型的契合程度 Can be used with any large model, this version is more suitable for the vision, the whole body, of course, depending on your preferences, the current test if you want to only want a little effect but want to have an impact object I suggest that the Lora value is 0.2-0.4 so that may have a little effect, but not too strong, is still in the basic version, the follow-up may be more iterations and updates LoraI trained lora based on anything large model, so I think its fit strength is very high and I am currently testing the fit degree of Lora fit model

Chinese ink painting

<ul><li>Lora权重0.5-1更好，小于0.8＞0.5产生的融合底模型的效果更好</li><li>Lora weight 0.5-1 is better, and fusion bottom models less than 0.8 &gt; 0.5 produce better results</li></ul>

Files with SHA256: ab9d943106ccb2f5617d6bddee5890817619c0fa1c467bafc0ae397c0a5fec46