Purple Dreamy Flux Lo R A

Introduction

Purple-Dreamy-Flux-LoRA is a text-to-image model developed by prithivMLmods. It utilizes the LoRA (Low-Rank Adaptation) technique in conjunction with diffusion models to generate images based on textual descriptions. The model is still in the training phase and may not be fully optimized.

Architecture

• Base Model: The model is built on black-forest-labs/FLUX.1-dev.
• LoRA Technique: This model applies LoRA, a method to adapt large models efficiently with fewer parameters.
• Image Processing: Uses AdamW optimizer with a constant learning rate scheduler and specific noise parameters to enhance image quality.

Training

• Parameters:
  • Learning Rate Scheduler: Constant
  • Optimizer: AdamW
  • Network Dimensions: 64
  • Alpha: 32
  • Epochs: 15
• Data: Trained on a dataset of 17 high-resolution images using the florence2-en labeling system.
• Output Dimensions: Best images produced at 1024 x 1024 resolution.

Guide: Running Locally

1. Setup Environment:

   • Ensure you have torch and pipelines libraries installed.
   • Use a cloud GPU for better performance, such as those from AWS or Google Cloud.

2. 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/Purple-Dreamy-Flux-LoRA"
   trigger_word = "Purple Dreamy"  
   pipe.load_lora_weights(lora_repo)
   
   device = torch.device("cuda")
   pipe.to(device)
   

3. Generate Images:

   • Use the trigger word "Purple Dreamy" to produce images using the pipeline.

License

The Purple-Dreamy-Flux-LoRA model is licensed under CreativeML OpenRAIL-M, which allows for creative use while ensuring responsible AI deployment.