ferferefer/lora_stable_diffusion

LoRA Model

Model Description

A LoRA fine-tune of the Anime (animefull-final-pruned) model.

Base Model

This LoRA was trained on animefull-final-pruned-fp16.safetensors

Training Details

• Type: LoRA
• Base Model: animefull-final-pruned-fp16
• Resolution: 512
• Epochs: 1
• Learning Rate: 5e-4 (unet), 1e-4 (text encoder)
• Rank (Network Dimension): 16
• Alpha: 8
• Batch Size: 2
• Optimizer: AdamW8bit
• LR Scheduler: cosine_with_restarts (3 cycles)
• LR Warmup: 5% of total steps

Usage

To use this LoRA with Automatic1111's WebUI:

1. Install the base model (animefull-final-pruned-fp16)
2. Place the LoRA file in the models/Lora directory
3. In the WebUI, add the LoRA to your prompt with <lora:model_name:1.0>

Examples

[Add your example images and prompts here]

License

[Specify your preferred license]

Training Configuration Details

• Flip Augmentation: Disabled
• Min SNR Gamma: Enabled (5.0)
• Mixed Precision: fp16
• Clip Skip: 2
• Network Module: networks.lora
• XFormers: Enabled
• Caption Extension: .txt

Acknowledgments

[Any acknowledgments or credits you want to add]