这是一个简单的概念Lora，它能够较好的生成出手杖而不是一堆乱七八糟的钢管或是黑色线条（当然，可能仍然需要碰运气）。不过请注意，它偶尔会生成两根手杖，让你的人物看起来像是在滑雪，以及可能会让你的画面变得灰暗——我暂时没有找到解决这个问题的方法。另外还有一点小问题，我在测试时发现在Counterfeit-V2.5模型上使用高清修复有较大概率生成纯黑图片。我不清楚其他模型是否会出现这类问题。The following is AI translation:It's a simple concept Lora, which can generate a cane rather than a bunch of messy steel pipes or black lines (although it may still require some luck). However, please note that it occasionally generates two canes, making your character look like they are skiing and may make your screen look dark - I haven’t found a solution to this problem yet.In addition, there is a small problem. During the test, I found that using high-definition restoration on the Counterfeit-V2.5 model has a higher probability of generating a pure black picture. I don't know if other models have this kind of problem.

Cane

进行了更多次数的训练，现在它出现手杖的概率更高了此外，正在尝试训练一个Lycoris以获取更好的效果另外好奇如何使用正则化修复这个模型总是会让角色变成滑雪运动员的问题，希望能够给出建议（Trained more times, now it has a higher chance of showing the caneAlso, trying to train a Lycoris for better resultsIn addition, I am curious about how to use regularization to fix the problem that this model always makes the character become a skier, hoping to give suggestions (

Files with SHA256: 36ff0f44f10c72a16ea1089632a96de818fa7c6b00bd6f4595786acb3c2dfa8c