Neural-Siege is a PyTorch GPU multi-agent arena sim (thousands of agents) with PPO training, checkpoint/resume, and detailed telemetry. I’m using it to probe emergent behavior and failure modes (camping/stalemates). Would appreciate critiques on the environme…
A formal, reproducible, and extensible multi-agent grid simulation framework for large-scale reinforcement-learning experiments, telemetry-driven analysis, and long-running headless execution.
This … [+14588 chars]
