creator: FallKiller
I have encountered GPU video memory problems in the creation of Lora and checkpoint. The 4060 8G I currently use cannot train pictures with resolutions above 768x768, which greatly increases my training time, and 8G video memory cannot train checkpoint. So I've posted a request for donations here on a website: https://ko-fi.com/zun666
Here, I would like to thank the people who donated me!!!
国内支付宝、微信用户可以通过爱发电支持我:https://afdian.net/a/999qx killer (usells.com)
Enter trigger words; detached sleeves,gloves,hair ornament,japanese clothes,killer,pelvic curtain,wide sleeves
weight is recommended to use around 0.6-0.85! More of my models :killer (usells.com)
Local or cloud map parameter configuration recommended:
Sampling method Schedule type: DPM++ SDE Karras or DPM++ 2M SDE Karras or Euler
Number of iteration Steps: 30
Resolution Size: 768x768 or 720-1280
Open HD Repair √
我的更多美女服装Lora在吐司主页:(5414) Creator profile killer | 吐司 tusi.cn
我的更多美女服装Lora在libu主页:LiblibAI·哩布哩布AI - 中国领先的AI创作平台
点开我的头像主页看更多劲爆LORA!!!
qq交流群;815380848(已满可以进2群)
qq模型反馈交流二群:590377791
接Lora模型定制+q:1983331501 (备注来意!打包批量下载我训练的几百个美女服饰模型也可加)
输入触发词;detached sleeves,gloves,hair ornament,japanese clothes,killer,pelvic curtain,wide sleeves
权重推荐0.6-0.85左右即可使用!
本地或者云端跑图参数配置推荐:
采样方法Schedule type:DPM++ SDE Karras 或者DPM++ 2M SDE Karras或者Euler
迭代步数Steps:30
分辨率Size:768x768或者720-1280
开启高清修复√