Donate Coffee for Gtonerov1.1 implemented two EMA momentum calculations. Strongly sensitive to learning, responding better to images used But the results are still being tested.ip_noise has been turned off because it is not suitable for training anime images. make outline thinner Therefore it is compatible with every model.Main tag : slingshot swimsuitSecondary tag : wedgie, pubic hair,Extra Tag : clothing aside, Color tag : color slingshot swimsuit, LoRA str : 1Denoising str : 0.3Image Train : 1007Training Pony LoRA Powershell info :resolution = 768Batch size = 3network_dim = 16network alpha = 16mix/save_precision = BF16optimizer_type = AdEMAMixoptimizer_args_other = [Default optimizer]caption_tag_dropout_rate = 0.2 #Reduce over-reliance on prompts.GPU RTX 4060Ti 16GBTorch 2.3+cu121 + Xformers + gradient_checkpointingVram used = 11.1GBSpeed = 2.03sec/itWeightUNet weight average strength: 0.009330498257718395Text Encoder (1) weight average strength: 0.008645846338885195Text Encoder (2) weight average strength: 0.006813761007413265
