GenRL is a PyTorch-based reinforcement learning library designed for researchers and practitioners seeking reproducible and generalizable algorithm implementations. It aims to accelerate RL research by simplifying the reproduction of state-of-the-art algorithms and facilitating benchmarking.

How It Works

GenRL employs a PyTorch-first, modular architecture, enabling extensibility and idiomatic Python usage. It provides ready-made implementations of popular RL algorithms, a unified trainer and logging class for code reusability, and tools for faster benchmarking, including automated hyperparameter tuning and environment implementations. The library is structured to support new algorithm implementations in under 100 lines of code.

Quick Start & Requirements

• Install: pip install genrl
• Prerequisites: Python 3.6+, PyTorch, OpenAI Gym.
• Upgrade: pip install -U genrl
• From Source: git clone https://github.com/SforAiDl/genrl.git && cd genrl && python setup.py install
• Documentation: https://genrl.readthedocs.io/en/latest/

Highlighted Details

• Offers 20+ tutorials covering basic to state-of-the-art Deep RL algorithms.
• Implements a wide range of algorithms including DQN variants, PPO, SAC, TD3, and classical methods like Q-Learning and SARSA.
• Supports various Bandit RL algorithms, including contextual bandits and deep contextual bandits.
• Includes features for parameter noise sampling and Bayesian regression for posterior inference.

Maintenance & Community

• Active development with releases on PyPI (latest v0.0.2).
• Community support available via Slack: https://join.slack.com/t/genrlworkspace/shared_invite/zt-gwlgnymd-Pw3TYC~0XDLy6VQDml22zg
• CI/CD pipelines for Linux, macOS, and Windows.

Licensing & Compatibility

• License: MIT License.
• Compatibility: Permissive license suitable for commercial use and integration with closed-source projects.

Limitations & Caveats

The library is at version v0.0.2, and the README explicitly states to "Expect breaking changes." While extensive, the breadth of algorithms may mean some implementations are less mature or optimized than dedicated libraries.