Trained on Illustrious-XL.
ddkk:
Interestingly, I trained a style LoRA using a dataset based on this X user's works – the catch is that his art is also AI-generated. Despite this, I still named the style after them. Pretty meta, right?
Prompt :
masterpiece, best quality, amazing quality, Negative Prompt :
bad quality, worst quality, worst detail, sketch, censor, optional negative prompt :
monochrome, greyscale, signature, url, artist name, username, text, English text,Description
<p>{</p><p> "engine": "kohya",</p><p> "unetLR": 1,</p><p> "clipSkip": 2,</p><p> "loraType": "lora",</p><p> "keepTokens": 1,</p><p> "networkDim": 16,</p><p> "numRepeats": 7,</p><p> "resolution": 1984,</p><p> "lrScheduler": "cosine",</p><p> "minSnrGamma": 5,</p><p> "noiseOffset": 0.1,</p><p> "targetSteps": 3945,</p><p> "enableBucket": true,</p><p> "networkAlpha": 8,</p><p> "optimizerType": "Prodigy",</p><p> "textEncoderLR": 1,</p><p> "maxTrainEpochs": 46,</p><p> "shuffleCaption": true,</p><p> "trainBatchSize": 4,</p><p> "flipAugmentation": false,</p><p> "lrSchedulerNumCycles": 3</p><p>}</p>