Flux Creative Stocks LoRA

🚀 Flux-Creative-Stocks-LoRA

This project is a LoRA model based on the black-forest-labs/FLUX.1-dev base model, used for text-to-image generation. It can generate various creative stock images according to specific text prompts.

🚀 Quick Start

Prerequisites

Before using this model, you need to ensure that the following libraries are installed:

import torch
from pipelines import DiffusionPipeline

Loading the Model

base_model = "black-forest-labs/FLUX.1-dev"
pipe = DiffusionPipeline.from_pretrained(base_model, torch_dtype=torch.bfloat16)

lora_repo = "strangerzonehf/Flux-Creative-Stocks-LoRA"
trigger_word = "Creative Stocks"  
pipe.load_lora_weights(lora_repo)

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

Generating Images

You should use Creative Stocks to trigger the image generation. For example:

prompt = "Creative Stocks, A beautiful landscape with mountains and a lake"
image = pipe(prompt).images[0]
image.show()

✨ Features

• Text-to-Image Generation: Generate various creative stock images according to text prompts.
• LoRA Model: Based on the black-forest-labs/FLUX.1-dev base model, it can be easily integrated into existing pipelines.

📦 Installation

To use this model, you need to install the following dependencies:

pip install diffusers torch

💻 Usage Examples

Basic Usage

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 = "strangerzonehf/Flux-Creative-Stocks-LoRA"
trigger_word = "Creative Stocks"  
pipe.load_lora_weights(lora_repo)

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

prompt = "Creative Stocks, A flat lay of four neatly folded blue fabric rolls arranged in a square pattern. Each roll has a white label reading 'COMFY-COTTON' with a minimalist logo of a cotton plant. The backdrop is a soft beige fabric."
image = pipe(prompt).images[0]
image.show()

Advanced Usage

You can adjust the inference parameters to get different image effects:

# Adjust inference steps
num_inference_steps = 35
image = pipe(prompt, num_inference_steps=num_inference_steps).images[0]
image.show()

📚 Documentation

Image Processing Parameters

Labeling

Labeling: florence2-en(natural language & English)

Total Images Used for Training

Total Images Used for Training : 24 [ Raw 8-bit ]

Best Dimensions & Inference

Inference Range

• Recommended Inference Steps: 30–35

🔧 Technical Details

This model is a LoRA model based on the black-forest-labs/FLUX.1-dev base model. It uses the DiffusionPipeline for image generation. During training, 24 raw 8-bit images were used, and the labeling was done using the florence2-en natural language model in English.

📄 License

This project is licensed under the apache-2.0 license.