[Illustrious] Yuuki Sakuna (結城さくな) (2 Costume)

https://civitai.com/models/1146561/illustrious-yuuki-sakuna-2-costume

Yuuki Sakuna (結城さくな) - Illustrious LoRA (2 Costume)DisclaimerDo not use my LoRA to produce AI image and tagging her fanart hashtag in Twitter/X or if you have an enough contribution in that image then fine.NO REPOST IN ANY WEBSITE (PixAI, Tensor . Art, Seaart, ...) !!!!!!!!Version 3 descriptionNew Dataset and captions2nd costume is comingCompletely change any costumeThrow LoCon welcome UNet + CLIP-L (maybe future I will fork lycoris to support UNet + Conv + CLIP-L only)Completely use only her costume to cosplay by another character.Trigger Word (V3)default costumeyuuki sakuna, yuuki sakuna (1st costume), 1girl, long hair, blush, animal ears, pink hair, medium breasts, hair ornament, maid headdress, bow, maid, animal ear fluff, pink bow, puffy sleeves, apron, white dress, pink dress, blue ribbon, bowtie, collarbone, hairclip, hair bow, detached collar, white thighhighs, black footwear, shoes, black choker 2nd costume (maybe you can add some prompt for better output) some parts still missing due lack of dataset.yuuki sakuna, yuuki sakuna (2nd costume), 1girl, cat ears, short hair, pink hair, hair ribbon, pink eyes, choker, dress, white dress, purple dress, hair ornament, hairclip, bandaid hair ornament, long sleeves, bow, frills, white apron, handbag, wings any costumeyuuki sakuna, 1girl, long hair, blush, cat ears, pink hair, medium breasts, hair ornament, blue ribbon, hairclip, hair bow hairstyles (you can select ones)long hair, two sides up short hair single braidLimitationsSecond Costume outputs still not good enough.Training Details (V3)Network Modulestandard LoRA with clip-l onlyDataset1st costume 313 imagesalternative costume 63 images2nd costume (short hair) 17 images2nd costume (single braid) 12 imagesspecial dataset for unlocking ability to change any costume 50 imagesParametersresolution = 1024batch size = 2dim,alpha = 16/16mix/save precision = bf16/bf16optmizer = CAME + weight_decay=0.01UNet LR = 3.2e-5TE LR = 1.6e-5scheduler = cosine_with_min_lr (down to 0.25 ratio)alpha maskblock weights LRl2 lossStepsepochs = 10total steps = 5580repeats = 1,5,10,14,3Toolscustom kohya-ss GUItorch 2.7.0 cu128 + xformersRTX 3060 12 GBAverage WeightsUNet average weights = 0.00975080994381146TE1 average weights = 0.008148421067744493PREVIOUS VERSIONVersion 2 descriptionBig Improvement when changing costume (almost 99% can change into any costume except maid)New Datasetusing Lycoris LoCon instead of LoRAVersion 1 descriptionFirst Illustrious LoRAusing Lion SchedulerTrigger Worddefault costumeyuuki sakuna, long hair, blush, animal ears, pink hair, medium breasts, hair ornament, maid headdress, bow, maid, animal ear fluff, pink bow, puffy sleeves, apron, white dress, pink dress, blue ribbon, bowtie, collarbone, hairclip, hair bow, detached collar, white thighhighs, black footwear, shoes, black chokerremove black footwear, shoes, black choker, white thighhighs for half bodyany costumeyuuki sakuna, long hair, blush, cat ears, pink hair, medium breasts, hair ornament, blue ribbon, hairclip, hair bowLimitationsnothing (if found please comment below.)Training Details (V2)LoRA methodLycoris LoConDataset256 imagesparametersresolution = 1024batch size = 2dim,alpha = 16,12ConvDim, ConvAlpha = 16,12mix/save precision = bf16/bf16optmizer = Lion + weight_decay=0.1UNet LR = 4e-5TE LR = 1e-5scheduler = cosine_with_min_lrl2 lossstepsepochs = 10total steps = 2560repeats = 2toolskohya-ss GUI v24.3.0torch 2.5.0 cu124Runpod RTX 4090 24 GB + xformers + gradient_checkpointingweightUNet average weights : 0.009442168148234486TE1 average weights : 0.007911427761428058TE2 average weights : 0.0064397353562526405*This LoRA is for studying LoRA training with new technique so do not use for damaging the vtuber (also support her too).*REPOST model is not allow if you found this model please report to me.