LLuminate

LLuminate is an evolutionary computation framework that evolves artifacts using generative AI. It enables running experiments with various parameters to guide the evolutionary process.

Installation

# Clone the repository
git clone https://github.com/joel-simon/lluminate.git
cd lluminate

# Install dependencies
pip install -r requirements.txt

macOS Apple Silicon Support

This project supports Apple Silicon (M1/M2/M3) Macs using Metal Performance Shaders (MPS). The code will automatically detect and use MPS acceleration when running on Apple Silicon hardware. No additional configuration is needed.

When you run the application, you should see:

Using PyTorch device: mps

Note: On Apple Silicon, Metal Performance Shaders provide significant speedups for tensor operations compared to CPU execution.

Renderers Installation

To install certain renderers, cd to one of the subdirectories and run npm install:

cd src/render-shaders && npm install
cd src/render-sdf && npm install
cd src/render-p5js && npm install

System Requirements for Renderers

The renderers require some system dependencies. On macOS, install them using Homebrew:

brew install pkg-config cairo pango libpng jpeg giflib librsvg pixman

Note: Node.js 18.x is required for the renderers to work properly.

Usage

Run an evolutionary experiment using the main.py script:

python main.py --prompt "Your creative prompt here" --artifact_class shader --num_generations 20

Command Line Arguments

Argument	Description	Default
--output_name	Name of output directory	Current timestamp
--random_seed	Random seed for reproducibility	42
--prompt	Prompt for the experiment	""
--initial_population_size	Size of the initial population	20
--population_size	Size of the population in each generation	20
--children_per_generation	Number of children to create per generation	10
--num_generations	Number of generations to run	20
--k_neighbors	Number of neighbors for selection	3
--max_workers	Maximum number of parallel workers	5
--artifact_class	Class of artifact to evolve	"shader"
--evolution_mode	Mode of evolution	"variation"
--reasoning_effort	Level of reasoning effort	"low"
--no_strategies	Disable use of creative strategies	False
--no_summary	Disable summary usage	False
--crossover_rate	Probability of crossover during reproduction	0.3

Example

# Run an evolution experiment with a custom prompt and 30 generations
python main.py --prompt "Create a sunset shader with mountain silhouettes" --num_generations 30 --reasoning_effort high

Project Structure

lluminate/
├── main.py                 # Main entry point for running experiments
├── src/
│   ├── run_evolution_experiment.py  # Implementation of the evolution experiment
│   └── ...                 # Other source files
├── results/                # Generated artifacts and experiment results
└── ...

Output

Results are saved in the results/ directory, organized by artifact class and timestamp (or custom name if provided). Each experiment creates its own directory with:

• Generated artifacts for each generation
• Logs of the evolutionary process
• Summary of experiment parameters and results

License

Apache 2.0