Detailed Background

This is my first test. The training data was quite small, only 2 versions of 15 images, and only anime and furry was used so I don't know how well it will work for other uses.

Training hyperparameters:

{
  "unetLR": 0.0005,
  "clipSkip": 1,
  "loraType": "lora",
  "keepTokens": 0,
  "networkDim": 32,
  "numRepeats": 2,
  "resolution": 1024,
  "lrScheduler": "cosine_with_restarts",
  "minSnrGamma": 5,
  "noiseOffset": 0.1,
  "targetSteps": 600,
  "enableBucket": true,
  "networkAlpha": 16,
  "optimizerType": "AdamW8Bit",
  "textEncoderLR": 0.00005,
  "maxTrainEpochs": 40,
  "shuffleCaption": true,
  "trainBatchSize": 4,
  "flipAugmentation": false,
  "lrSchedulerNumCycles": 3
}