A model for large breasted waifus or semi-realistic characters.V6 Changelog 2023/06/03:<ul><li>Considering this was my first and most popular LoRA, I figured it was time to improve on it. Probably the last version too. Hyperfusion is more or less the successor to this.</li><li>Less harmful to the overall style, but slightly smaller sizes overall, feels like a good balance</li><li>Wider variety of breast shapes in the dataset<ul><li>Some breast shape tags were added, but hard to say how helpful they are</li></ul></li><li>Uses LoRA LoCon, but does not require any additional extensions to run</li><li>Changed "large breasts" to "big breasts" to match my hyperfusion model tags</li><li>The model prefers resolutions around 640, but can do up to 768 to a degree</li><li>Training tags attached under model download section</li></ul>V5 Changelog:<ul><li>v5 now uses LoRA!</li><li>What changed since v4? I trained a breast size classifier to auto label the dataset which resulted in better control over breast size, and made it easier to generate larger sizes.</li><li>Note that the trigger words in v5 are ordered in the following from smallest size to largest: breasts, large breasts, huge breasts, gigantic breasts</li></ul>V4 Info:the keyword for v4 remains hyperbreasts, if you want better control over the size use v5Notes:I used this to train my image tagging classifiers for breast sizes <a target="_blank" rel="ugc" href="https://github.com/huggingface/transformers/tree/main/examples/pytorch/image-classification">https://github.com/huggingface/transformers/tree/main/examples/pytorch/image-classification</a>

hyperbreasts

<ul><li>Added some more variety in breast shapes to the dataset</li><li>Less harmful to the overall style than v5</li></ul>Training Details:<ul><li>2k images</li><li>16 epocs</li></ul><ul><li>learning rate 2e-4</li></ul><ul><li>text encoder LR 5e-5</li><li>base model Av3</li><li>clip skip 2</li><li>random flip</li><li>tag drop chance 0.1</li><li>bucketing at 768</li><li>dim 16</li><li>alpha 8</li><li>150 tokens</li><li>cosine with restarts</li><li>training with tags, tags attached next to model download</li><li>Kohya LoRA LoCon, does not require any additional extensions to use</li></ul>

Files with SHA256: 40de0c12fbfc7aaf2195189ec2672a51ced79573d16d11b7a05dd7ae797a37f6