ファティマ-XLPony Fatima からくりサーカス/Karakuri Circusからくりサーカスを久々に読んでゲームがファティマ大好きな石油王に支えられてたという話を思い出したということでGPT-4oによるタグ付けのテストがてらSSRファティマの画像を主として学習させた藤田和日郎の書いた漫画は一枚も使ってないウェイトが高すぎると主線が太くなりすぎるので若干強度を落として使用するのを推奨Having revisited Karakuri Circus after a while, I remembered the story about the game being supported by an oil tycoon who loved Fatima.Thus, as a test of tagging by GPT-4o, I mainly trained on images of SSR Fatima.No manga drawn by Kazuhiro Fujita was used.It's recommended to slightly reduce the weight as too high a weight can make the main lines too thick.

ファティマ-XLPony Fatima  からくりサーカス/Karakuri Circus

<pre><code>bucket_no_upscale = true
bucket_reso_steps = 64
cache_latents = true
cache_latents_to_disk = true
caption_dropout_rate = 0.05
epoch = 30
gradient_accumulation_steps = 1
gradient_checkpointing = true
huber_c = 0.1
huber_schedule = "snr"
keep_tokens = 3
loss_type = "l2"
lr_scheduler = "cosine"
lr_scheduler_num_cycles = 1
lr_scheduler_power = 1
max_grad_norm = 1
max_timestep = 1000
min_snr_gamma = 5
mixed_precision = "bf16"
network_alpha = 1
network_dim = 16
no_half_vae = true
optimizer_args = [ "decouple=True", "weight_decay=0.01", "d_coef=0.8", "use_bias_correction=True", "safeguard_warmup=False", "betas=0.9,0.99",]
optimizer_type = "Prodigy"
</code></pre>

Files with SHA256: fa5074e910a1d38974e32f3f377ba03bb4774857d138e239dbcc86bdd9c65834