<h2 id="heading-414">Usage</h2><ul><li><p>Activation tag: <code>wakana shiki</code> </p></li><li><p>Additional tags: <code>blue hair, short hair, brown eyes, bangs</code> </p></li><li><p>Recommended LoRA ratio: 0.7-0.8</p></li></ul><h2 id="heading-415">Training details</h2><ul><li><p>Dataset: 203 anime screenshots </p></li><li><p>Reference training configuration:</p><pre><code>[additional_network_arguments]
unet_lr = 0.00025
text_encoder_lr = 0.0001
network_dim = 32
network_alpha = 16
network_module = "networks.lora"

[optimizer_arguments]
learning_rate = 0.00025
lr_scheduler = "cosine_with_restarts"
lr_scheduler_num_cycles = 3
lr_warmup_steps = 213
optimizer_type = "AdamW8bit"

[training_arguments]
max_train_epochs = 14
save_every_n_epochs = 2
save_last_n_epochs = 10
train_batch_size = 2
clip_skip = 2
min_snr_gamma = 5.0
weighted_captions = false
seed = 42
max_token_length = 225
xformers = true
lowram = true
max_data_loader_n_workers = 8
persistent_data_loader_workers = true
save_precision = "fp16"
mixed_precision = "fp16"
output_dir = ""
logging_dir = ""
output_name = "shiki"
log_prefix = "shiki"
save_state = false

[model_arguments]
pretrained_model_name_or_path = "/content/AnyLoRA_noVae_fp16-pruned.ckpt"
v2 = false

[saving_arguments]
save_model_as = "safetensors"

[dreambooth_arguments]
prior_loss_weight = 1.0

[dataset_arguments]
cache_latents = true

% ---------

[[datasets]]

[[datasets.subsets]]
num_repeats = 3
image_dir = "/content/drive/MyDrive/Loras/shiki/dataset"

[general]
resolution = 512
shuffle_caption = true
keep_tokens = 1
flip_aug = false
caption_extension = ".txt"
enable_bucket = true
bucket_reso_steps = 64
bucket_no_upscale = false
min_bucket_reso = 256
max_bucket_reso = 1024</code></pre></li></ul>

Files with SHA256: 03a2c1364cfc382d282e7c418bb3e176ac4f4e87aced58bcaf57a093d88ecfc8