哈哈，我常常会训练一些，有趣的但是莫名其妙的lora模型，这次是在推特上看到许多人喜欢这样的脏鞋子，所以就训练了这个模型。它是在 FLUX.DEV完整模型上使用FP16的精度训练的，同时将clip-L模型换成了CLIP-GmP-ViT-L-14 模型，然后使用了joy_caption_two进行自然语言训练集打标。95张训练集中，包含尖头高跟鞋，平底鞋，系带凉鞋，厚底高跟鞋，等等常见的鞋类，但是有点遗憾，我忘了添加一些皮革长靴进去。我没有设置触发词，你只需要直接了当的输入你想要的画面即可。Haha, I often train some interesting but inexplicable lora models. This time I saw many people like such dirty shoes on Twitter, so I trained this model.It is trained with FP16 accuracy on the <a target="_blank" rel="ugc" href="http://FLUX.DEV">FLUX.DEV</a> full model, and the clip-L model is replaced with the CLIP-GmP-ViT-L-14 model, and then joy_caption_two is used for natural language training set labeling.The 95 training sets include pointed high heels, flat shoes, lace-up sandals, thick-soled high heels, and other common footwear, but it’s a bit regrettable that I forgot to add some leather boots.I didn’t set a trigger word, you just need to directly input the picture you want.

Files with SHA256: 14f80bfb32f380cd21dce3458805b947c674d0682acff09d63e31d751602203a