Try experimenting within the 3–10 range until you find the output that best fits your preference.
Recommended Model:
Recommended workflow:
For best results, use Text-to-image Hi-res fix (denoise around: 0.3) to enhance details and overall quality.”
Recommended Use Settings:
- Lora Weight: 1 , CFG Scale: 3-10 , Sampler: Euler a , Steps: 30 and higher
- ADetailer for Face Fix is highly recommended.
- image-to-image Hi-res fix, (Also perfect for face fix)
- Img2img Upscale ( Remacri )
Description
Training Parameters:
Dataset Images: 53 Images
Epochs: 10
Num Repeats: 8
Train Batch Size: 4
Steps: 1060
Resolution: 1024
Enable Bucket: True
Shuffle Tags: True
Keep Tokens: 1
Clip Skip: 1
Flip Augmentation: False
Unet LR: 0.0001
Text Encoder LR: 0.00005
LR Scheduler: cosine_with_restarts
LR Scheduler Cycles: 3
Min SNR Gamma: 5
Network Dim: 32
Network Alpha: 16
Noise Offset: 0.1
Optimizer: Adafactor