Paired Open-Ended Trailblazer (POET) is an open-ended reinforcement learning algorithm designed to continuously generate increasingly complex and diverse learning environments and their solutions. It targets researchers and practitioners in AI seeking to overcome the limitations of traditional RL benchmarks by fostering unbounded invention and adaptation.

How It Works

POET employs a population-based approach where agents and environments co-evolve. New environments are generated by perturbing existing ones, and agents are trained to solve these evolving challenges. This creates a positive feedback loop, driving increasing complexity and diversity in both the environments and the learned policies.

Quick Start & Requirements

• Install: Clone the repository and install dependencies.
• Prerequisites: Fiber, NEAT-Python, OpenAI Gym.
• Local Run: ./run_poet_local.sh final_test
• Kubernetes Run: Follow instructions for persistent volume setup and use ./run_poet_remote.sh final_test.
• Logs: fiber cp nfs:/persistent/logs/final_test . and fiber cp nfs:/persistent/logs/poet_final_test poet_final_test.
• Cluster Deployment: Refer to Fiber Documentation for containerization and cluster execution.

Highlighted Details

• Implements both POET and Enhanced POET algorithms.
• Focuses on open-endedness and the invention of learning challenges.
• Supports local execution and distributed training on Kubernetes via Fiber.

Maintenance & Community

The project is associated with Uber Engineering. Further community or roadmap details are not explicitly provided in the README.

Licensing & Compatibility

The README does not specify a license. Compatibility for commercial or closed-source use is not detailed.

Limitations & Caveats

The project relies on proprietary Fiber infrastructure for distributed execution, which may limit its portability. The "legacy branch" is mentioned for the original POET, suggesting potential ongoing development or divergence between versions.