📌 General Information
LoRA Name: Odette
Base Model: Stable Diffusion XL v1.0 (sdxl_base_v1-0)
Training Resolution: 1024x1024
Training Steps: 510
Epochs: 10
Dataset Size: 208 training images
Training Date: 2024-07-23
🔧 Training Configuration
LoRA Network Dimensions: dim=32, alpha=32
Max Token Length: 225
Learning Rate: 1.0 (both UNet and text encoder)
Batch Size per Device: 5
Gradient Checkpointing: Enabled
Mixed Precision: bf16
Optimizer: Prodigy with weight decay 0.5 and betas=(0.9, 0.99)
Learning Rate Scheduler: cosine
Clip Skip: 1
Cache Latents: True
Caption Dropout Rate: 0.0
🏷️ Training Tags (with frequency)
From the ss_tag_frequency and dataset annotations:
👤 Character and Body
1girl (26), solo (22), breasts (3), large breasts (1), breasts out (1)
1boy (3), ass (1), nipples (2), penis (2), pubic hair (1)
💇♀️ Hair & Eyes
blonde hair (25), long hair (26), blue eyes (12), purple eyes (3)
👗 Clothing & Accessories
dress (15), white dress (1), gown (1), puffy sleeves (9), juliet sleeves (1), necklace (3), jewelry (4)
💄 Makeup
lipstick (13), red lips (3), makeup (9)
🧍♀️ Pose & Expression
smile (6), open mouth (3), looking to the side (1), crossed arms (1), one eye closed (1)
🎭 Style & Art
retro artstyle (7), style parody (1), flat color (1), 1980s (style) (1), parody (3), portrait (1)
🌄 Scene & Background
sky (4), tree (2), curtains (1), table (1), playing card (1), card (1), night (1), outdoors (1), flower (1), simple background (1), white background (1)
⚠️ NSFW Tags
uncensored (1), sex (1), hetero (2)