LoRA概念
LoRA的全称是LoRA: Low-Rank Adaptation of Large Language Models，可以理解为stable diffusion（SD)模型的一种插件，和hyper-network，controlNet一样，都是在不修改SD模型的前提下，利用少量数据训练出一种画风/IP/人物，实现定制化需求，所需的训练资源比训练SD模要小很多，非常适合社区使用者和个人开发者。
下载模型
基础模型
chilloutmix_Ni.safetensors官方版下载丨最新版下载丨绿色版下载丨APP下载-123云盘
cp ~/Downloads/ai/chilloutmix_Ni.safetensors models/Stable-diffusion/chilloutmix_Ni.safetensorsLoRA模型
yeqiang@yeqiang-MS-7B23:~/Downloads/ai/stable-diffusion-webui$ ls models/Lora/ chunmomoLora_v11.safetensors HinaIAmYoung22_zny10.safetensors raidenShogunRealistic_raidenshogun.safetensors taiwanDollLikeness_v10.safetensors yorBriarSpyFamily_lykonV1.safetensors hf2ming-beta-020601.safetensors koreanDollLikeness_v15.safetensors styleJelly_v10.safetensors yaeMikoRealistic_yaemikoMixed.safetensors角色模型
到参考资料链接下载所有模型
出图
正向提示词
best quality, ultra high res, (photorealistic:1.4), 1girl, full body, <lora:koreanDollLikeness_v15:1> ,<lora:yaeMikoRealistic_yaemikoMixed:1>,反向提示词
paintings, sketches, (worst quality:2),(low quality:2),(normal quality:2), lowres, normal quality, ((monochrome)), ((grayscale)), skin spots, acnes, skin blemishes, age spot, glans,{{{bad_hands}}},badhand, bad_feet,故障
OutOfMemoryError
OutOfMemoryError: CUDA out of memory. Tried to allocate 450.00 MiB (GPU 0; 5.79 GiB total capacity; 3.13 GiB already allocated; 422.62 MiB free; 3.60 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
Time taken: 33.77s
Torch active/reserved: 3948/4000 MiB, Sys VRAM: 5796/5927 MiB (97.79%)
显卡内存顶不住了，缩小图片尺寸。说明：显存大小比显卡性能重要，显存不够直接崩溃，显卡性能差点顶多慢点。
Failed to match keys when loading Lora
原来这玩意是基础模型，把这种基础模型移动到models/Stable-diffusion/ 目录下
mv models/Lora/chilloutmix_Ni.safetensors models/Stable-diffusion/模型目录状态
在图形界面上选择该基础模型
参考资料
什么是LoRA模型，如何使用和训练LoRA模型？你想要的都在这！ - 知乎
LORA模型使用教程及模型下载 Stable Diffusion WebUI - openAI
codediffusion基础模型utmwebuistable diffusionpytorch提示词ide参考资料webopenaidoc个人开发者图片尺寸unreal图形界面定制化chromecontrolnetragapp数据训练gpu开发者url
