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 }
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: