这是一款需要使用特定tag才能发挥作用的lora模型，请使用以下的tag。This is a lora model that requires the use of a specific tag to function, please use the tag below.<img src="https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/977d4cbe-97c2-4098-871d-a6e92a01a800/width=525/977d4cbe-97c2-4098-871d-a6e92a01a800" />实验lora模型能承载多少细节的测试产物，非常好用的多姿势、多服饰生成器Experiment lora model can carry how much detail of the test product,very good multi-pose, multi-dress generator建议使用强度——0.2-0.4Recommended strength - 0.2-0.4使用额外的tag激活模型的特定方向Use additional tags to activate a specific direction of the model在真人模型中表现不佳Poor performance in live models特定的tag：解释说明afterwards——after sexarched_climax——climax wite head uplift——lift shirt，lift shortsreverse sweater——Wearing Backless sweater backwardsair_oral——Cue oral sex<img src="https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/6e593d90-1940-437c-c768-48274655b700/width=525/6e593d90-1940-437c-c768-48274655b700" />我认为lora的能力被低估了，lora可以学习到的概念不止一种，如果有足够的素材，应该可以得到一款多模型可用的强大lora模型。I think the ability of lora is underestimated. lora can learn more than one concept, and with enough material, it should be possible to get a powerful lora model with multiple models available.批量更改训练集的txt，能让lora学习到更准确的细节。Batch changing the txt of the training set enables lora to learn more accurate details.先在小批量的数据集上训练，不断加入新的数据，并在之前的模型上作迁移训练，这样收敛更快。Train on small batches of datasets first, keep adding new data, and do migration training on the previous model so that the convergence is faster.

Files with SHA256: d852d923957e5f6ce120cf82351dfa571b1fad145158f7bc1227334ffd53f7c1