v.1.1 &lt;== Stable version with "Prodigy" optimizer. You can use any weight without risk to get any kitties died ;)This version is just 2nd epoch, but already shows significant effect on picture. Feel free to try and comment.PS: "RiP Keith Flint"PPS: You can show your appreciation ==&gt; <a target="_blank" rel="ugc" href="https://www.buymeacoffee.com/zemskov">https://www.buymeacoffee.com/zemskov</a>v.1.0 &lt;== !!!AcHTUNG!!! Very UNsTabLE experimental Lora. Read Description CaReFuLlY! I am serious!Weight &lt;0.195: PANDORA 8==&gt; CAN'T EXPLAIN WHAT IT DOES EXACTLY ¯\_(ツ)_/¯ Complex improvement of cinematic look, subject separation (depth of field, swirly bokeh, etc), scenery mood (fog, flame/smoke epicness and realism, etc), anatomy (better poses, shyness, face / body expression, etc). Weight 0.195...0.21: Wipes out subject, but renders cool scenery with narrow depth of field and lovely creamy bokeh :PWeight &gt; 0.21: Explodes planets, stars, galaxies and initiates the spontaneous vacuum collapse at quantum level (GO THAT FAR IN ANOTHER UNIVERSE PLEASE :)

CinEro XL - PANDORA | LoCon / LoHa

I changed the training mode (now LoHa) and changed the training parameters in Kohya SS to fix the instability of optimizer. Optimizer also changed/ Now I use Prodigy.Full settings below: number of images (including repeats) / 各bucketの画像枚数（繰り返し回数を含む）bucket 0: resolution (768, 512), count: 816bucket 1: resolution (768, 896), count: 4bucket 2: resolution (768, 1024), count: 992bucket 3: resolution (768, 1152), count: 2170bucket 4: resolution (896, 768), count: 23bucket 5: resolution (896, 896), count: 71bucket 6: resolution (896, 1024), count: 98bucket 7: resolution (896, 1152), count: 40bucket 8: resolution (1024, 768), count: 1490bucket 9: resolution (1024, 896), count: 99bucket 10: resolution (1024, 1024), count: 802bucket 11: resolution (1152, 768), count: 763bucket 12: resolution (1280, 768), count: 540mean ar error (without repeats): 0.025714800356640224preparing acceleratorloading model for process 0/1load StableDiffusion checkpoint: /mnt/SD/.ckpt/SDXL/juggernautXL_version2.safetensorsUsing rank adaptation algo: lohaUse Dropout value: 0.0Create LyCORIS ModuleCreate LyCORIS Modulecreate LyCORIS for Text Encoder: 264 modules.Create LyCORIS Modulecreate LyCORIS for U-Net: 788 modules.enable LyCORIS for U-Netprepare optimizer, data loader etc.Deprecated: use prepare_optimizer_params(text_encoder_lr, unet_lr, learning_rate) instead of prepare_optimizer_params(text_encoder_lr, unet_lr)use Prodigy optimizer | {}override steps. steps for 40 epochs is / 指定エポックまでのステップ数: 63280enable full bf16 training.running training / 学習開始 num train images * repeats / 学習画像の数×繰り返し回数: 3954 num reg images / 正則化画像の数: 30064 num batches per epoch / 1epochのバッチ数: 7908 num epochs / epoch数: 40 batch size per device / バッチサイズ: 1 gradient accumulation steps / 勾配を合計するステップ数 = 5 total optimization steps / 学習ステップ数: 63280

Files with SHA256: 316969550c6efd4e7bce07ba7f19eb63b2310aaa20c18506d4e0f231a54efeb0