If you support me, please buy me a ko-fi.<a target="_blank" rel="ugc" href="https://ko-fi.com/turugini">https://ko-fi.com/turugini</a>This is a Lora studied only with images in which nothing is worn directly under the sailor collar or with swimwear. I think it is possible to specify the color to some extent in the prompts. Due to the convenience of the training image, a cannon may be attached to the image. Please specify this by using a negative prompt.In version 2, the learning method was changed from kohya_ss to TrainTrain.I finally understand stratified learning. I think this will remedy the problem of not being able to use Lora together.version 3 The learning method was changed back from TrainTrain to kohya_ss.Also,the model used for training was changed.

Files with SHA256: ba421111182bbc710cc16284775565ca137f93e24112748ae0433c5d203b5a71