Ton618 Tarot Cards Flux Lo R A

Introduction

The TON618-Tarot-Cards-Flux-LoRA model by prithivMLmods is designed for generating images in the style of tarot cards using a text-to-image approach. It is equipped with diffusion and LoRA techniques, allowing for creative and thematic visualizations based on textual prompts.

Architecture

The model employs the DiffusionPipeline, initially built on the "black-forest-labs/FLUX.1-dev" base model. It utilizes LoRA weights to enhance image generation capabilities. The architecture is optimized with parameters such as AdamW optimizer, a constant learning rate scheduler, and multiresolution noise handling.

Training

The model was trained using 17 high-resolution images. Key training parameters include an AdamW optimizer, a constant learning rate, noise offset of 0.03, and 21 repeat steps over 2000 iterations per epoch. The network dimension is set to 64, with an alpha of 32, and training data labels are based on natural language English descriptions.

Guide: Running Locally

1. Install Required Libraries:

   • Ensure that Python and PyTorch are installed on your machine.
   • Install the diffusers library.

   pip install torch diffusers
   

2. Set Up the Model:

   • Import necessary libraries and load the base diffusion 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)
   

3. Load LoRA Weights:

   • Download and load the LoRA weights to enhance the model's capabilities.

   lora_repo = "prithivMLmods/Ton618-Tarot-Cards-Flux-LoRA"
   pipe.load_lora_weights(lora_repo)
   

4. Run on GPU:

   • Set the device to CUDA for improved performance.

   device = torch.device("cuda")
   pipe.to(device)
   

5. Generate Images:

   • Use the trigger word ("Tarot card") to generate images.

   For enhanced performance, consider using cloud-based GPU services such as AWS, Google Cloud, or Azure.

License

The model is distributed under the CreativeML OpenRAIL-M license, which allows for creative and responsible use while adhering to specific community agreements.