If you’re looking to dive into the world of AI-driven image generation, the Stable Diffusion Latent Space Explorer is a fantastic tool. This article will guide you through the setup process and provide instructions for running experiments. We’ll also add some handy troubleshooting tips along the way!

General Information

Created by Alen Smajic under the guidance of Prof. Dr. Visvanathan Ramesh, this project is funded by Goethe University and the AI Systems Engineering Lab (AISEL).

Setup Instructions

Follow these steps to set up your environment:

1. Clone the repository using the command:

git clone https://github.com/alen-smajic/Stable-Diffusion-Latent-Space-Explorer

2. Create a virtual environment and activate it:

python -m venv venv

3. Install PyTorch with CUDA following the instructions in the PyTorch installation guide.
4. Install the required libraries:

pip install diffusers[torch] transformers

5. Optional: for efficient attention, install xFormers:

pip install xformers

Tested Versions

For optimal performance, make sure you are using the following versions:

• torch: 2.0.0
• cuda: 11.8
• diffusers: 0.14.0
• xformers: 0.0.16

Downloading Model Weights

You can download the model weights using git-lfs:

git lfs install

git clone https://huggingface.co/stabilityai/stable-diffusion-2-1

This command will create a local folder called .stable-diffusion-2-1 on your disk.

Running Experiments

To run an experiment, you will define experiment parameters through a configuration file before executing the experiment with the script:

python run_sd_experiment.py --exp_config .configs/experiments/path_to_your_config

Understanding the Code with an Analogy

Think of experimenting with Stable Diffusion as hosting a creative cooking contest. Each configuration file is akin to a recipe that tells you exactly how much of each ingredient (parameters) to use. When you run the script, it’s like putting your ingredients together and following the recipe. The results folder produced after your cooking session is like a kitchen that keeps track of all your dishes (experiment results) by storing the variations you created (the generated images), the recipes you used (the configuration files), and even those attempts that didn’t go as planned (error logs).

Troubleshooting

In case you encounter issues during any setup or experimental run, here are some troubleshooting tips:

• Ensure that your packages’ versions match those tested above. Mismatched versions can lead to unexpected behaviors.
• If the experiments don’t run as expected, double-check the paths provided to the configuration files. A common oversight is incorrect file paths!
• Remember that the results may vary even if you use the same configuration and random seed due to the inherent randomness of AI models.

For more insights, updates, or to collaborate on AI development projects, stay connected with fxis.ai.

Final Notes

At fxis.ai, we believe that such advancements are crucial for the future of AI, as they enable more comprehensive and effective solutions. Our team is continually exploring new methodologies to push the envelope in artificial intelligence, ensuring that our clients benefit from the latest technological innovations.