noobai X L E P S cyberfixv2 perpendicularcyberfixv2

Introduction

NOOBAI-XL-EPS-CYBERFIXV2-PERPENDICULARCYBERFIXV2 is an advanced text-to-image generative model developed using diffusion techniques. It is fine-tuned from the Laxhar/noobai-XL-1.0 model and utilizes extensive datasets for training, including Danbooru and e621, with native tag captions.

Architecture

The model is a diffusion-based text-to-image generative model fine-tuned from Laxhar/noobai-XL-1.0. It leverages cyber realistic v4 methodologies and EPS models, employing advanced techniques like sd_mecha for enhancements. The model operates with a CFG (Classifier Free Guidance) scale of 5 to 6 and uses Euler a as the sampling method. It supports multiple resolutions with a total area around 1024x1024 pixels.

Training

The training datasets include the latest Danbooru images and e621 datasets, ensuring high-quality image generation. Quality and date tags are applied to classify images based on popularity and recency. Training involves data normalization, time-based decay coefficients, and comprehensive ranking.

Guide: Running Locally

1. Environment Setup:

   • Install Python and required libraries, including PyTorch and Hugging Face Transformers.
   • Clone the repository from Hugging Face.

2. Model Loading:

   • Load the model using the Hugging Face transformers library.

3. Inference:

   • Use the provided prompt guidelines to generate images. Adjust settings like resolution and sampling method as needed.

4. Hardware Recommendations:

   • For optimal performance, use cloud GPUs such as those provided by AWS, Google Cloud, or Azure.

License

The model is licensed under the fair-ai-public-license-1.0-sd with additional terms. Usage restrictions include prohibitions on harmful activities, offensive content generation, and commercialization. Derivative works must be open-sourced, and users must comply with open-source community guidelines. Users assume all risks related to model outputs.