模型/CAT - Citron Styles - 真實風格動畫

CAT - Citron Styles - 真實風格動畫

|
8/3/2025
|
3:16:59 AM
| Discussion|
0
被稱為索倫之塔的暗黑黑塔,頂部有巨大發光紅眼,周圍是火山山脈,流動的熔岩與火紅天空下的火山口荒原。
停滯罐內呈胎兒姿勢懸浮的小型外星怪物特寫,帶柔和陰影及簡單背景。
一位成熟女性肖像,栗色長髮綁成寬鬆馬尾,開心地笑著,背景是晴朗藍天並有明顯飛機尾跡,穿著色彩繽紛的花卉印花校服泳衣。
穿著校服、長髮女孩輪廓,站在被自然環繞的石頭小徑上,背後是明亮的斑駁陽光與散景效果。

推薦提示詞

score_9, score_8_up, score_8, (ultra realistic, 32k, masterpiece:1.2), medium breasts, (high detailed skin:1.1), ( high quality:1.1), (curvy), cute, eyelashes, blush, looking at viewer

推薦反向提示詞

score_6, score_5, score_4, busty, ugly face, mutated hands, low res, blurry face, black and white

推薦參數

samplers

DPM++ 2M Karras, DPM++ 2M, Euler a

steps

30

cfg

7

clip skip

2

resolution

832x1216, 832x1248

版本亮點

我在觸發詞裡加了一些建議的提示,但其實你不一定需要使用它們。

--

你可以在我多個 Lora 發布貼裡找到部分訓練數據,比如這個 https://civitai.com/images/8452326

--

感謝 iJWiTGS8 幫我學會如何創建這個風格!

--

訓練資料量: 210 張圖片

步驟數: 840

總時間: 2:36:04

--

[network_arguments]

unet_lr = 0.75

text_encoder_lr = 0.75

network_dim = 32

network_alpha = 32

network_module = "networks.lora"

network_train_unet_only = false

[optimizer_arguments]

learning_rate = 0.75

lr_scheduler = "cosine_with_restarts"

lr_scheduler_num_cycles = 3

lr_warmup_steps = 42

optimizer_type = "Prodigy"

optimizer_args = [ "weight_decay=0.1", "betas=[0.9,0.99]",]

[training_arguments]

pretrained_model_name_or_path = "stablediffusionapi/pony-diffusion-v6-xl"

vae = "stabilityai/sdxl-vae"

max_train_epochs = 10

train_batch_size = 5

seed = 42

max_token_length = 225

xformers = false

sdpa = true

min_snr_gamma = 8.0

lowram = true

no_half_vae = true

gradient_checkpointing = true

gradient_accumulation_steps = 1

max_data_loader_n_workers = 8

persistent_data_loader_workers = true

mixed_precision = "fp16"

full_bf16 = false

cache_latents = true

cache_latents_to_disk = true

cache_text_encoder_outputs = false

min_timestep = 0

max_timestep = 1000

prior_loss_weight = 1.0

[saving_arguments]

save_precision = "fp16"

save_model_as = "safetensors"

save_every_n_epochs = 1

save_last_n_epochs = 10

output_name = "RealisticAnime"

output_dir = "/content/drive/MyDrive/lora_training/output/RealisticAnime"

log_prefix = "RealisticAnime"

logging_dir = "/content/drive/MyDrive/lora_training/log"

創作者贊助

上一個
Vibrantly Sharp style - Vibrantly Dark style v.4
下一個
Grunge 字體設計 - v1.0

模型詳情

模型類型

LORA

基礎模型

Pony

模型版本

Realistic Anime

模型雜湊值

8b5dc12fc3

訓練詞彙

(ultra realistic, 32k, masterpiece:1.2), (high detailed skin:1.1), (high quality:1.1),
bokeh, luminescent background,

創作者

討論

log in以發表評論。