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.

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——0.2-0.4

Recommended 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 sex

arched_climax——climax wite head up

lift——lift shirt，lift shorts

reverse sweater——Wearing Backless sweater backwards

air_oral——Cue oral sex

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.