3 D Render Flux Lo R A

Introduction

The 3D-Render-Flux-LoRA model is a text-to-image generation model specialized for creating 3D portraits and renders. It integrates with the Diffusers library and utilizes LoRA (Low-Rank Adaptation) techniques to generate detailed and stylistic images based on text prompts.

Architecture

The model employs a base model named black-forest-labs/FLUX.1-dev and is enhanced with LoRA weights to support specialized 3D rendering. The architecture is designed to interpret descriptive text prompts and generate corresponding high-quality 3D images.

Training

The model is currently undergoing training, using a total of 19 high-resolution images. Key training parameters include:

• LR Scheduler: constant
• Optimizer: AdamW
• Noise Offset: 0.03
• Network Dimensions: 64
• Epochs: 15 with saves every epoch

The model is tuned to perform image generation using specific trigger words like "3D Portrait" and "3D render."

Guide: Running Locally

1. Setup:

   • Ensure you have PyTorch installed with CUDA support for GPU acceleration.
   • Install the necessary libraries such as diffusers.

2. Code Snippet:

   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/3D-Render-Flux-LoRA"
   trigger_word = "3D Portrait, 3d render"  
   pipe.load_lora_weights(lora_repo)
   
   device = torch.device("cuda")
   pipe.to(device)
   

3. Execution:

   • Use the trigger words "3D Portrait, 3d render" to generate images.

4. Recommended Cloud GPUs:

   • Consider using cloud services like AWS EC2 with NVIDIA GPUs, Google Cloud Platform, or Azure for optimal performance.

License

The model is distributed under the CreativeML Open RAIL-M license, which permits open use in compliance with the license terms.