| Property | Value |
|---|---|
| License | CreativeML OpenRAIL-M |
| Base Model | black-forest-labs/FLUX.1-dev |
| Training Images | 13 |
| Network Architecture | LoRA (64 dim, 32 alpha) |
What is Flux.1-Dev-Sketch-Card-LoRA?
Flux.1-Dev-Sketch-Card-LoRA is a specialized fine-tuning model designed for generating sketch card artwork. Built on the FLUX.1-dev base model, it utilizes Low-Rank Adaptation (LoRA) technology to create detailed sketch card illustrations with minimal training data. The model is currently in development phase, trained on a carefully curated dataset of 13 images.
Implementation Details
The model implements a constant learning rate scheduler with AdamW optimizer, featuring a network dimension of 64 and alpha of 32. It incorporates advanced noise handling with a 0.03 offset and multires noise iterations set to 10. The training process spans 16 epochs with 1990 steps per repeat cycle.
- Optimal resolution: 768 x 1024 (Best performance)
- Alternative resolution: 1024 x 1024 (Default)
- Florence2-en labeling system for natural language processing
- Implements noise discount of 0.1 for enhanced detail generation
Core Capabilities
- Generation of detailed sketch card artwork
- Handles complex character illustrations
- Supports various artistic styles and compositions
- Maintains consistent quality across different subjects
Frequently Asked Questions
Q: What makes this model unique?
This model specializes in sketch card generation with minimal training data (13 images) while maintaining high quality output through optimized LoRA parameters and noise handling techniques.
Q: What are the recommended use cases?
The model is ideal for creating character-based sketch cards, cartoon illustrations, and stylized artwork. It performs best with the trigger word "sketch card" and at the recommended 768x1024 resolution.