Trained on a dataset of 58 images (x 10 repeats) with manually reviewed WD14 captions and with regularisation images against sdXL_v10VAEFix model
Kohya settings:
Optimizer: Prodigy
LR/TE: 1
Sheduler: Cosine
Train batch size: 1
Res: 1024,1024
Trained on a dataset of 58 images (x 10 repeats) with manually reviewed WD14 captions and with regularisation images against sdXL_v10VAEFix model
Kohya settings:
Optimizer: Prodigy
LR/TE: 1
Sheduler: Cosine
Train batch size: 1
Res: 1024,1024