Canopus Lo R A Flux Ultra Realism 2.0

Introduction

The Canopus-LoRA-Flux-UltraRealism-2.0 model is a text-to-image model utilizing LoRA (Low-Rank Adaptation for Transformers) fine-tuning techniques to produce hyper-realistic images. This model is part of the Hugging Face ecosystem and is designed to enhance photorealism in image generation tasks.

Architecture

• Base Model: Uses black-forest-labs/FLUX.1-dev as the foundation.
• Model Type: LoRA, specialized in image style transfer and realism.
• Parameters:
  • LR Scheduler: Constant
  • Optimizer: AdamW
  • Network Dimensions: 64
  • Network Alpha: 32
  • Epochs: 20 with save checkpoints every epoch.
  • Noise Offset: 0.03
  • Multires Noise Discount: 0.1
  • Multires Noise Iterations: 10
  • Repeat & Steps: 30 & 3.8K+

Training

• Training Images: Utilized 70 high-resolution images.
• Labeling: Done with florence2-en for natural language and English.
• Current Status: The model is still in the training phase and may exhibit artifacts or suboptimal performance.

Guide: Running Locally

To run the Canopus-LoRA-Flux-UltraRealism-2.0 model locally:

1. Setup Environment: Ensure you have Python and necessary libraries installed, including PyTorch and Hugging Face Diffusers.
2. Download Model: Retrieve the model weights in Safetensors format from the Hugging Face Files & Versions tab.
3. Install Dependencies:

   pip install torch diffusers
   

4. Load Model:

   import torch
   from pipelines import DiffusionPipeline
   
   base_model = "black-forest-labs/FLUX.1-dev"
   pipe = DiffusionPipeline.from_pretrained(base_model, torch_dtype=torch.bfloat16)
   
   lora_repo = "prithivMLmods/Canopus-LoRA-Flux-UltraRealism-2.0"
   pipe.load_lora_weights(lora_repo)
   
   device = torch.device("cuda")
   pipe.to(device)
   

5. Run Inference: Use the pipeline to generate images with the trigger word "Ultra realistic".

Cloud GPUs

For enhanced performance, especially for high-resolution image generation, utilizing cloud GPUs from providers like AWS, Google Cloud, or Azure is recommended.

License

The model is distributed under the creativeml-openrail-m license. For specific licensing information, refer to the Hugging Face model page.