CivChan was created based on the dataset distributed by LoRA.Animagine4, iLLustrious-v1.0, LoRA.New-Dataset adds 34 images to the original dataset above.The entire tag has been re-created and re-structured, and two types of captions have been added Three types of tags were given.Caption: JoyCap-alpha2, Florence2-Large,Tags: JoyTAG, WDv3, e621Training: TrainTrain-LoRA, 1536pix, RAdamSF, 4e-04, 230step, D16A16,If you are interested, please try out <a target="_blank" rel="ugc" href="https://github.com/hako-mikan/sd-webui-traintrain">TrainTrain</a> by referring to the LoRA metadata.TrainTrain can create not only LoRA but also useful sliders, etc. <a target="_blank" rel="ugc" href="https://civarchive.com/articles/12334/new-lora-training-method-addiftalternating-direct-difference-training">ADDifT</a> (Alternating Direct Difference Training), an advanced version of Slider (Difference),The simplified version of LECO, <a target="_blank" rel="ugc" href="https://github.com/hako-mikan/sd-webui-traintrain/blob/main/README.md">i-LECO</a>, is also useful, so please try it.<hr />CivChan LoRA で配布されているデータセットをもとに作成しました。Animagine4、iLLustrious-v1.0、のLoRAです。New-Dataset は上記の元のデータセットに 34画像を付け足し、全てのタグを以下の内容で再作成＆再構成し、キャプションは２種類、タグは３種類を付与しました。キャプション：JoyCap-alpha2、Florence2-Large、タグ：JoyTAG、WDv3、e621トレーニング：TrainTrain-LoRA、1536pix、RAdamSF、4e-04、230step、D16A16、興味のある方は LoRA のメタデータを参考に <a target="_blank" rel="ugc" href="https://github.com/hako-mikan/sd-webui-traintrain">TrainTrain</a> をお試しくださいTrainTrain は LoRA だけでなく便利なスライダー等もつくれますSlider (Difference) の発展型 <a target="_blank" rel="ugc" href="https://civarchive.com/articles/12334/new-lora-training-method-addiftalternating-direct-difference-training">ADDifT</a> (Alternating Direct Difference Training) や、簡易版 LECO の <a target="_blank" rel="ugc" href="https://github.com/hako-mikan/sd-webui-traintrain/blob/main/README.md">i-LECO</a> も便利です、こちらもお試しください

Files with SHA256: 07917711338a39365834ec89e837e83c9bef01fb15b8bdaf026da6ff70d7de74