Petri Clash

Two neural cellular automata grow, collide, and regenerate on a shared grid.

• v2: Living arena: the actively improved version, with mouse-only lab tools, guided regrowth, saveable seeded duels, paired comparisons, territory/pressure views, and CPU-first play. Includes trained weights; no GPU or training is needed to try it.
• v1: the original, self-contained baseline. Unchanged by the v2 upgrade.

Quick start (Python 3.11+):

On Apple silicon, use python -m pip install torch instead of the CPU-index command below. For other platforms, the official PyTorch installer can select a compatible wheel.

python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate
python -m pip install torch --index-url https://download.pytorch.org/whl/cpu
python -m pip install -r v2/requirements.txt
python v2/clash.py --device cpu

The two versions do not share code, targets, or weights. See the verification reports for measured results and limitations.