我训练这个lora的目的是因为我想到了一个全新的一图流，或者叫少图流的训练思路（感谢reo以及L_A_X特供的灵感）之后我会将极少数据集训练出来的lora发布出来与使用正常数量数据集的lora进行对比如果你喜欢，请点个赞吧（The reason I trained this loraBecause I came up with a new idea for training a one-graph flow, or less-graph flow (thanks to reo and L_A_X for the inspiration)Later I will publish the lora trained on a very small number of datasetsCompare it with the lora using a normal number of datasetsIf you like it, please give it a like (

Azure_(Alchemy_stars) 

新的训练流程，仅使用4张图片作为数据集但是效果比传统方法训练出来的几乎没有差别，在某些地方上甚至表现的更好教程将会在最近几天发布到b站New training process, using only 4 images as datasetHowever, the results are almost indistinguishable from traditional training methods, and in some places even better.The tutorial will be posted on the bilibili in the next few days

Files with SHA256: f9fe58f6ce319dc11bd6499331874dedda4722c9c0300b956ed9ae4458e7dcb2