Flux Gym: Easy UI for training FLUX LoRA with low video memory

General Introduction

Flux Gym is an easy web UI for training FLUX LoRA with low graphics memory support (12GB/16GB/20GB). The front-end is based on AI-Toolkit's Gradio UI, and the back-end is powered by Kohya Scripts.Flux Gym combines the simplicity of the AI-Toolkit WebUI with the flexibility of Kohya Scripts for a wide range of graphics memory configurations, and supports Docker and automatic model download features.

Function List

• Supports 12GB, 16GB, 20GB video memory
• Docker support with automatic model download
• Support for customized base models
• Automatic generation of sample images
• Posted to Huggingface
• Advanced Function Options (hidden)

Using Help

Installation process

one-click installation

1. utilizationPinokio one-click starterAutomatically installs and launches everything: Pinokio One-Click Launcher

manual installation

1. Clone Fluxgym and kohya-ss/sd-scripts:

   git clone https://github.com/cocktailpeanut/fluxgym
   cd fluxgym
   git clone -b sd3 https://github.com/kohya-ss/sd-scripts
   

2. Activate the virtual environment:
   • Windows.

     python -m venv env
     env\Scripts\activate
     

   • Linux.

     python -m venv env
     source env/bin/activate
     

3. Install the dependencies:

   cd sd-scripts
   pip install -r requirements.txt
   cd .
   pip install --r requirements.txt
   pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
   

4. Launch the application:

   python app.py
   

Docker Installation

1. Clone Fluxgym and kohya-ss/sd-scripts:

   git clone https://github.com/cocktailpeanut/fluxgym
   cd fluxgym
   git clone -b sd3 https://github.com/kohya-ss/sd-scripts
   

2. Build the image and run it:

   docker compose up -d --build
   

3. Open your browser and visit: http://localhost:7860

Functional operation flow

1. Enter the LoRA information.
2. Upload images and add tags (using trigger words).
3. Click on the "Start" button.

Sample Image Configuration

By default, Fluxgym does not generate sample images during training. You can configure Fluxgym to generate sample images at every N steps:

• Sample Image Cues: These cues will be used to automatically generate images during training.
• Sample images per N steps: For example, if the "Expected training steps" is 960 and the "Sample images per N steps" is 100, the images will be generated at the 100th, 200th and 300th steps.

Advanced Sample Images

Using the built-in Kohya sd-scripts syntax, you have full control over the sample images generated during the training phase:

• Trigger words: e.g., hrld person.
• Advanced flags: e.g., the --d flag specifies the seed, --w specifies the image width, --h specifies the image height, etc.

Posted to Huggingface

1. Get Huggingface Token:Huggingface Token
2. Enter the Token and click "Login".
3. Select the trained LoRA, edit the name and publish it to Huggingface.

The layman's guide to getting started with AI

Help you learn how to utilize AI tools at a low cost and from a zero base.AI, like office software, is an essential skill for everyone. Mastering AI will give you an edge in your job search and half the effort in your future work and studies.

View Details>