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)

now you can change the hair color, and nsfw

Ares_v2.safetensors

v1.0

v2.1

v1.25

v0.2

v0.1a

Plz add "3d_rendering" in ng tag to lower the 3D feeling Update the dataset: added hair color tag to allow the hair color to changeadded nsfw data for nsfw picUpdate of the training:edited the block weightsnetwork_args = [ "down_lr_weight=1,1,1,1,1,1,0.1,0.1,1,1,1,1", "mid_lr_weight=0.1", "up_lr_weight=1,1,1,1,1,1,1,1,1,1,0.1,0.1", "block_lr_zero_threshold=0" ]

Files with SHA256: 9eaa0ee6effb505155ea5593cdeb8669a0a8d4b992d76e89d4c7624675529fd3