GPTS2 | Lazy Style LoRA | Extended LoRA

<p>Second Lazy Style Lora. This time much bigger. It's still focused on anime-like style, so, difference from GPTS1 is not very big.</p><p>It is made on the base of extra lazy tagging way, which involves only barebones editing of auto-tagged images. This dataset consists of 404 images.100% SFW.<br />Additional experimental tag filtering was applied(maybe i'll release that project in future).</p><p>Of course, main theme of dataset is anime waifus, but considering dataset size, there are bunch of other themes to ease normalization of style to other themes, which led to better cooperation with tags that were not trained.</p><p>Generally, this is not supposed to be used for anythin complex, because it was not tagged in a way to be, but you still can try and show me your results :3<br /><br />Same as GPTS1 Lite - it will have a lot of bleed, but i noticed that it's less in this one.</p><p>If you want it to be editable and still have some style effect, try weight from 0.5 to 0.75. That should make it editable and still be looking good :D</p><p>Generally, this dataset processing used my custom tag filtering experiment, testing of which on GPTS1 led to increased editability, so, it might not be as hard to edit this one, but still, main purpose is just to get that bit of "wow" effect style.</p><p>Was not able to make it work with dynamic normalization, so, only rank 64(that did work for some reason, even though it's supposed to die with Adan d-adapt, as my experience with other trainings suggest). </p>