基于merchidark模型制作的wlop风格lora，爆肝72小时画素材，然后用自己的素材炼成,推荐使用0.8以上比重,如果需要hires放大，推荐重绘幅度0.3以下，请将adetail的重绘幅度改为0.2以下,NSFW测试通过，但可能会出现面部不够美的现象，至于其他，请自己摸索,
A LORA (Low-Rank Adaptation) model trained in WLOP's art style, based on the MerchiDark model. I spent 72 intense hours creating training materials/images, then used my materials to train the model. It is recommended to use a weight of 0.8 or higher. If you need to use hires upscaling, the Recommended denoising strength is 0.3. Please reduce the ADetail denoising strength to below 0.2. NSFW (Not Safe For Work) testing has been completed, though facial features may occasionally appear less attractive than desired. As for other aspects, please explore and experiment on your own.
描述:
用更细节的图像炼制，跟v1相比，可以随意设置，且已通过nsfw测试。无需adetailer或者低重绘率，hires重回幅度3.5，步数在50步以上，最好搭配meichidark模型使用。
Uses more detailed image training compared to v1, with flexible customizable settings, and has passed NSFW testing. No need for adetailer or low redraw rates. The hires upscale ratio is 3.5, with step count recommended above 50 steps. Best used in combination with the meichidark model.
训练词语:
名称: dewlop07-000126.safetensors
大小 (KB): 499030
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success
