FLUX LoRa

A LoRA adapter tailored to the FLUX.1-dev AI image model, with a focus on generating personalized images.

NOTE: To protect the privacy of the data, the LoRA weights as well as the data used to train the LoRA is not included in this repository.

Background

LoRAs

LoRAs are small, efficient adapters that can be added to a model to improve its performance on a specific task. They are typically used in conjunction with a base model, such as a large language model or an image model.

For more information, please read the LoRA paper.

FLUX.1-dev

FLUX.1-dev is a state-of-the-art AI image model that can generate high-quality images from text descriptions.

For more information, please read the official blog post from Black Forest Labs.

Sample Images

generated_images/1.jpg	generated_images/2.jpg

generated_images/3.jpg	generated_images/4.jpg

Training Your Own LoRA

Install Dependencies

First, you need to install the project dependencies using pip:

pip install -r requirements.txt

Gather Data

Then, in the /images directory, place the training images in .png format. You should have at least 12 high-quality images to train the LoRA. (Example: /images/1.png, /images/2.png, etc.)

Add captions

Also in the /images directory, you must place the captions for the images in .txt format. Each caption should be in a separate file and the file name should match the image file name. (Example: /images/1.txt, /images/2.txt, etc.)

Train LoRA

Finally, you can train the LoRA by running the following command:

python train.py --config train_configs/lora.yaml