(Flux) Tangled end credits sequence版本CivitAI Rapid Train (ID: 950297)

Kind of a silly idea, but this LoRA is trained to produce images that resemble the heavily stylized end-credit sequence of the Disney film Tangled.

V1.0 was trained on 80 images with the following parameters:

{

  "engine": "kohya",

  "unetLR": 0.0005,

  "clipSkip": 1,

  "loraType": "lora",

  "keepTokens": 0,

  "networkDim": 2,

  "numRepeats": 4,

  "resolution": 1024,

  "lrScheduler": "cosine_with_restarts",

  "minSnrGamma": 5,

  "noiseOffset": 0.1,

  "targetSteps": 1920,

  "enableBucket": true,

  "networkAlpha": 16,

  "optimizerType": "AdamW8Bit",

  "textEncoderLR": 0,

  "maxTrainEpochs": 12,

  "shuffleCaption": false,

  "trainBatchSize": 2,

  "flipAugmentation": false,

  "lrSchedulerNumCycles": 3

}

描述:

This was the output LoRA from CivitAI's rapid training. I don't believe the quality is quite as good as the one trained with my own parameter set, but figured I'd share it so people can compare.

训练词语: ink on paper