Uncoloured Polygon Flux Lo R A

Introduction

The Uncoloured-Polygon-Flux-LoRA model by prithivMLmods is designed for text-to-image generation using the LoRA (Low-Rank Adaptation) technique. It specializes in creating uncolored polygonal images with distinct features and minimalistic backgrounds, suitable for abstract and artistic interpretations.

Architecture

The model utilizes the black-forest-labs/FLUX.1-dev as its base model and employs various parameters for image processing:

• LR Scheduler: Constant
• Optimizer: AdamW
• Network Dim: 64
• Network Alpha: 32
• Repeat & Steps: 27 & 2000
• Epoch: 13

Additionally, noise parameters such as Noise Offset, Multires Noise Discount, and Multires Noise Iterations are used to refine the images.

Training

The model is still in the training phase and uses a set of over 25 high-resolution images. It incorporates florence2-en for natural language processing and English labeling. Training images are sourced from muz.li.

Guide: Running Locally

1. Clone the Repository: Fetch the model files from the Files & versions tab.
2. Install Dependencies: Ensure that you have the necessary Python packages, including those for machine learning and image processing.
3. Load the Model: Use a script to load the model weights in Safetensors format.
4. Generate Images: Use prompt-based inputs to generate images focusing on the trigger word "uncoloured polygon".

Cloud GPU Suggestion

For optimal performance, consider using cloud GPUs from providers like AWS, Google Cloud, or Azure to handle the computational load of model inference and training.

License

The model is licensed under creativeml-openrail-m, allowing for creative and research use under specified conditions.