SimpleTuner Docker

A Docker container for SimpleTuner, a general fine-tuning kit for diffusion models. This repository provides a pre-configured environment for training Stable Diffusion, SDXL, Flux, and other diffusion models.

Quick Start

Prerequisites

• Docker installed on your system
• Git installed on your system
• Basic familiarity with command line operations
• GPU with at least 8GB VRAM (16GB+ recommended)

First Time Setup

1. Clone this repository:

   git clone https://github.com/chrevdog/SimpleTuner.git
   cd SimpleTuner

2. Set up the upstream connection (one-time setup):

   git remote add upstream https://github.com/bghira/SimpleTuner.git

3. Build the Docker image:

   docker build -t theloupedevteam/simpletuner-docker:latest .

4. Run the container:

   docker run -it --gpus all -p 8888:8888 -v $(pwd)/storage:/workspace/storage theloupedevteam/simpletuner-docker:latest

5. Access JupyterLab:

   • Open your browser and go to http://localhost:8888
   • No password or token required

Updating SimpleTuner

When you want to update to the latest version of SimpleTuner:

1. Fetch Latest Changes

# Fetch the latest changes from upstream
git fetch upstream

# Switch to main branch
git checkout main

# Merge the latest changes
git merge upstream/main

2. Update the SimpleTuner Submodule

# Navigate to the SimpleTuner directory
cd SimpleTuner

# Pull the latest changes
git pull origin main

# Go back to the root directory
cd ..

3. Rebuild the Docker Image

# Build with the new version
docker build -t theloupedevteam/simpletuner-docker:latest .

# Tag with version number (optional but recommended)
docker tag theloupedevteam/simpletuner-docker:latest theloupedevteam/simpletuner-docker:v2.1.2

4. Push to Docker Hub

# Login to Docker Hub (if not already logged in)
docker login

# Push the latest version
docker push theloupedevteam/simpletuner-docker:latest

# Push the versioned tag
docker push theloupedevteam/simpletuner-docker:v2.1.2

5. Commit and Push Changes

# Add all changes
git add .

# Commit with a descriptive message
git commit -m "Update SimpleTuner to v2.1.2"

# Push to your repository
git push origin main

Training Configuration

Basic Training Command

Minimum Requirements

• RAM: 24 min, aim for 48 GB system RAM

Troubleshooting

Common Issues

1. Out of Memory (OOM) Errors:

   • Reduce batch size: --train_batch_size 1
   • Use LoRA: --use_lora
   • Enable mixed precision: --mixed_precision "fp16"
   • Use gradient accumulation: --gradient_accumulation_steps 4

2. Slow Training:

   • Increase batch size if memory allows
   • Use --mixed_precision "fp16"
   • Consider using multiple GPUs with DeepSpeed

3. Poor Training Results:

   • Check your dataset quality and captions
   • Adjust learning rate (try 1e-4 to 1e-5)
   • Increase training steps
   • Use proper data augmentation

Documentation

For detailed information about SimpleTuner features and capabilities, refer to:

• SimpleTuner Documentation
• Quickstart Guides
• Training Tutorials

Contributing

1. Fork this repository
2. Create a feature branch: git checkout -b feature-name
3. Make your changes
4. Commit your changes: git commit -m 'Add feature'
5. Push to the branch: git push origin feature-name
6. Submit a pull request

License

This project is licensed under the AGPL-3.0 License - see the LICENSE file for details.

Support

For issues related to:

• SimpleTuner functionality: Original SimpleTuner Issues
• Docker container: Create an issue in this repository
• Training problems: Check the SimpleTuner Documentation