What happens if we are able to put multiple different NCAs in a single substrate that compete for space?
curl -LsSf https://astral.sh/uv/install.sh | sh # Or install uv another way
git clone https://github.com/SakanaAI/petri-dish-nca
cd petri-dish-nca
uv sync
- basic:
uv run python src/train.py --n-ncas 3 --epochs 1000 --device cpu
- wandb logging:
uv run python src/train.py --n-ncas 3 --epochs 10000 --device cuda --wandb
For additional configurations, you can load a JSON config file.