I was trying to train my Au Ra in FF14, by providing screenshot from the game. Its can output the horn and scales right, but the 3D style is affecting the generation. Anyone have a idea of how to fix it?These are some of the settingpretrained_model_name_or_path = "./sd-models/latest.ckpt"resolution = "640,640"enable_bucket = truemin_bucket_reso = 256max_bucket_reso = 1024save_every_n_epochs = 1max_train_epochs = 10train_batch_size = 2network_train_unet_only = falsenetwork_train_text_encoder_only = falselearning_rate = 0.0001unet_lr = 3.5e-5text_encoder_lr = 3e-6lr_scheduler = "cosine_with_restarts"optimizer_type = "Lion"lr_scheduler_num_cycles = 1network_module = "networks.lora"network_dim = 32network_alpha = 32shuffle_caption = trueweighted_captions = falsekeep_tokens = 1max_token_length = 255seed = 1337prior_loss_weight = 1clip_skip = 2gradient_checkpointing = falsemixed_precision = "fp16"save_precision = "fp16"xformers = truelowram = falsecache_latents = truecache_latents_to_disk = falsepersistent_data_loader_workers = true

Aura (FFxiv)

I change the base model to AOM3 and added some face Lora during generation, might try to merge some face Lora into this on later on. Also will training Others Aura other then my own to share it.

Files with SHA256: 6da47368d047d5b97bd162d6677c2b5123da4b521ead88431ddaba3c5772a978