SDXL LoRA/LyCORIS works best on the model that it was trained on. I will release multiple versions for a few popular models. Feel free to request for artist or model.
Version naming convention: MajorVersion.MinorVersion [BaseModel]
MajorVesion: For big updates that can apply to any style LoRA. For example, training parameter update.
MinorVersion: For small updates that only apply to this LoRA. For example, epoch selection, removing a certain image from dataset.
BaseModel: SDXL LoRA/LyCORIS works best on the model that it was trained on.
Description
FAQ
Comments (11)
朋友你训练的画风lora质量都高的离谱啊，可以分享一下你的技巧和参数设置之类的吗
the quality of the art style you’ve trained with LoRA is exceptionally high. Could you share some of your techniques and parameter settings?
谢谢你的支持，LoRA训练就是不断尝试的过程，我现在每天都在尝试新的参数。
关于参数设置技巧，我推荐使用：
Prodigy optimizer
gradient accumulation
scale weight norms
每个LoRA具体的参数设置可以在它的metadata里查看到
@silvermoong 你好，我模仿了你的训练设置，因为我有244张训练集，所以我把batch size和gradient accumulation分别设置为2和61，epoch设置为100，scale weight norms设置为1，使用了prodigy optimizer，其他参数也大部分照搬你的其他lora，但是结果出现了明显的过拟合现象，这是什么原因，如果你帮助我，我将非常感激（Hello, I imitated your training settings. I have 244 training sets, so I set the batch size and gradient accumulation to 2 and 61 respectively, epoch to 100, scale weight norms to 1, and used prodigy optimizer. Most of the other parameters are copied from your other loras, but the result is obvious overfitting. What is the reason? If you help me, I will be very grateful.）
@3324304459321 你是指第100号epoch过拟合吗？试试看稍前一些的epoch
Generated images all use dpoxl lora?
nevermind sorry I saw you linked it somewhere else
@Aruhshura where
@Aruhshura tysm
请问这个dpoxl的lora是起到什么作用的呢