Open-source reinforcement-learning software
267 open-source libraries, tools and applications tagged reinforcement-learning, ranked by how closely reinforcement-learning is their primary purpose, then by adoption and maintenance health. Each entry links to a page with sources, use cases and when to choose or avoid it.
267 products, primary matches first, then adoption-weighted; health v2 shown.
| Product | Health v2 | Stars | Maturity |
|---|---|---|---|
| Godot RL Agents Godot RL Agents is an open-source Python package that bridges games built in the Godot Engine with reinforcement learning algorithms, enabl… | 60 | 1605 | active |
| Equim-chan/Mortal Mortal is a free and open-source AI for Japanese (riichi) mahjong powered by deep reinforcement learning, written in Rust with a Python int… | 48 | 1580 | active |
| modelscope/AgentEvolver AgentEvolver is an end-to-end self-evolving training framework for LLM-based agents that unifies automatic task generation (self-questionin… | 47 | 1571 | active |
| google-deepmind/bsuite bsuite (Behaviour Suite for Reinforcement Learning) is a collection of carefully-designed experiments from DeepMind that investigate core c… | 79 | 1564 | stable |
| RLHFlow/RLHF-Reward-Modeling A collection of training recipes for reward models used in RLHF, covering Bradley-Terry reward models, pairwise preference models, ArmoRM, … | 30 | 1546 | active |
| SERL HIL-SERL is a Python library suite for training reinforcement learning policies for precise robotic manipulation using human demonstrations… | 41 | 1533 | active |
| Leela Zero Leela Zero is an open-source Go engine that reimplements AlphaGo Zero, combining Monte Carlo Tree Search with a deep residual convolutional… | 23 | 5591 | maintenance |
| keras-rl/keras-rl A Python library implementing state-of-the-art deep reinforcement learning algorithms (DQN, DDPG, SARSA, and more) that integrates seamless… | 23 | 5549 | maintenance |
| sail-sg/envpool EnvPool is a C++-based batched environment pool with pybind11 bindings and a thread pool for high-performance parallel RL environment execu… | 96 | 1522 | active |
| open-thought/reasoning-gym Reasoning Gym is a Python library of procedural dataset generators and algorithmically verifiable reasoning environments for training LLMs … | 60 | 1512 | active |
| fan-ziqi/rl_sar A C++ framework for simulation verification and physical deployment of reinforcement learning policies for robots, supporting quadruped, wh… | 74 | 1481 | active |
| google-deepmind/rlax RLax is a JAX-based library of building blocks for implementing reinforcement learning agents, providing mathematical operations like value… | 79 | 1449 | active |
| microsoft/TextWorld TextWorld is a Python library from Microsoft that generates text-based adventure games and provides a sandbox learning environment for trai… | 72 | 1448 | active |
| Agent-RL/ReCall ReCall is a framework that trains LLMs to reason with arbitrary tool calls via reinforcement learning, without supervised tool-use trajecto… | 26 | 1441 | active |
| microsoft/rStar Microsoft's research repository for rStar2-Agent, a 14B math reasoning model trained with agentic reinforcement learning that autonomously … | 38 | 1427 | active |
| Denys88/rl_games RL Games is a high-performance reinforcement learning library built on PyTorch, focused on training agents in massively parallel GPU-based … | 75 | 1397 | active |
| unitreerobotics/unitree_rl_lab A collection of reinforcement learning environments for Unitree robots (Go2, H1, G1) built on NVIDIA IsaacLab. It enables training locomoti… | 53 | 1378 | active |
| erwincoumans/tiny-differentiable-simulator Tiny Differentiable Simulator (TDS) is a header-only C++ and CUDA physics library for rigid-body dynamics with zero dependencies, supportin… | 67 | 1376 | active |
| Goal-Driven Autonomous Exploration (DRL Robot Navigation) A ROS Gazebo simulation project that trains a mobile robot to navigate to random goals while avoiding obstacles using a TD3 deep reinforcem… | 54 | 1374 | active |
| kengz/SLM-Lab SLM Lab is a modular deep reinforcement learning framework built in PyTorch, offering ready-to-use algorithms like PPO, SAC, DQN, and A2C v… | 91 | 1365 | active |
| ZJU-REAL/ClawGUI ClawGUI is a unified Python framework for GUI agents covering the full lifecycle: online reinforcement learning training (ClawGUI-RL with G… | 64 | 1356 | active |
| jonathan-laurent/AlphaZero.jl A generic, simple, and fast Julia implementation of DeepMind's AlphaZero algorithm for training game-playing agents via self-play and MCTS.… | 70 | 1338 | active |
| CarperAI/trlx trlX is a distributed training framework for fine-tuning large language models with reinforcement learning from human feedback (RLHF), supp… | 23 | 4756 | maintenance |
| kwai/DouZero DouZero is a deep reinforcement learning framework that masters the Chinese card game DouDizhu through self-play, published at ICML 2021 by… | 23 | 4685 | maintenance |
| suragnair/alpha-zero-general A clean, flexible implementation of the AlphaZero self-play reinforcement learning algorithm that can be adapted to any two-player turn-bas… | 32 | 4520 | maintenance |
| ICT-FinD-Lab/alphagen AlphaGen is a Python research library that automatically generates formulaic alpha (predictive) stock factors using reinforcement learning,… | 67 | 1268 | active |
| metadriverse/metadrive MetaDrive is an open-source, lightweight driving simulator built for AI and autonomy research, supporting compositional scene synthesis and… | 68 | 1259 | active |
| NVIDIA-NeMo/Gym NeMo Gym is a Python library from NVIDIA for evaluating and improving LLM models and agents using environments. It provides infrastructure … | 81 | 1252 | active |
| google-deepmind/android_env AndroidEnv is a Python library from DeepMind that exposes an Android device (real or emulated) as a Reinforcement Learning environment. Age… | 83 | 1248 | active |
| rohanpsingh/LearningHumanoidWalking A Python research codebase for training humanoid robots to walk using deep reinforcement learning (PPO) in MuJoCo simulation. It provides e… | 66 | 1224 | active |
| microsoft/malmo Project Malmo is a platform for artificial intelligence experimentation and research built on top of Minecraft, providing a gym-like enviro… | 10 | 4271 | maintenance |
| JackHopkins/factorio-learning-environment An open-source framework for developing and evaluating LLM agents in the game of Factorio, providing an open-ended, non-saturating benchmar… | 83 | 1199 | active |
| chengxuxin/extreme-parkour Official code for 'Extreme Parkour with Legged Robots' (ICRA 2024), a reinforcement learning framework for training quadruped robots to per… | 28 | 1184 | active |
| PKU-Alignment/omnisafe OmniSafe is a PyTorch-based infrastructural framework for safe reinforcement learning research, providing a unified modular toolkit and com… | 24 | 1153 | active |
| nnaisense/evotorch EvoTorch is an open-source evolutionary computation library built on top of PyTorch, developed at NNAISENSE. It provides distribution-based… | 75 | 1148 | active |
| hanruihua/ir-sim IR-SIM is an open-source, Python-based lightweight robot simulator for navigation, control, and learning. It provides a simple YAML-driven … | 99 | 1146 | active |
| ZiwenZhuang/parkour Official code for 'Robot Parkour Learning' (CoRL 2023), a reinforcement learning system that trains quadrupedal robots to perform vision-ba… | 47 | 1133 | active |
| ikostrikov/pytorch-a2c-ppo-acktr-gail A PyTorch implementation of several deep reinforcement learning algorithms: A2C, PPO, ACKTR, and GAIL (imitation learning). It works with O… | 32 | 3907 | maintenance |
| lasgroup/SDPO SDPO (Self-Distilled Policy Optimization) is a research library implementing a reinforcement learning framework for post-training large lan… | 53 | 1126 | active |
| RUC-NLPIR/ARPO ARPO (Agentic Reinforced Policy Optimization) is a reinforcement learning algorithm and training framework for LLM agents, published at ICL… | 52 | 1126 | active |
| luchris429/purejaxrl PureJaxRL provides end-to-end reinforcement learning training pipelines implemented entirely in JAX, including environments, enabling massi… | 31 | 1118 | active |
| PKU-MARL/DexterousHands Bi-DexHands is a Python library providing bimanual dexterous hand manipulation environments built on NVIDIA Isaac Gym for reinforcement lea… | 32 | 1104 | active |
| Toni-SM/skrl skrl is an open-source modular Reinforcement Learning library written in Python, implemented in PyTorch, JAX, and NVIDIA Warp. It supports … | 86 | 1098 | active |
| LucasAlegre/sumo-rl SUMO-RL is a Python library that provides Gymnasium and PettingZoo-compatible reinforcement learning environments for traffic signal contro… | 51 | 1095 | active |
| agi-brain/xuance XuanCe is an open-source Python library of deep reinforcement learning (DRL) and multi-agent reinforcement learning (MARL) algorithm implem… | 91 | 1087 | active |
| alexfrom0815/Online-3D-BPP-PCT A Python research implementation of the ICLR 2022 paper 'Learning Efficient Online 3D Bin Packing on Packing Configuration Trees', using de… | 57 | 1087 | active |
| google-research/football Google Research Football is a reinforcement learning environment based on the open-source Gameplay Football game, created by the Google Bra… | 10 | 3669 | maintenance |
| danijar/dreamerv2 A TensorFlow 2 implementation of the DreamerV2 model-based reinforcement learning agent that learns world models from high-dimensional imag… | 32 | 1062 | stable |
| TeleHuman/PBHC Official implementation of KungfuBot, a physics-based humanoid whole-body control framework for learning highly-dynamic skills like kungfu … | 57 | 1061 | active |
| openai/retro Gym Retro is a Python library that turns classic video games into Gym environments for reinforcement learning research, with integrations f… | 10 | 3581 | maintenance |
| RLCard RLCard is a Python toolkit for reinforcement learning research in card games, providing environments for Blackjack, Leduc Hold'em, Texas Ho… | 23 | 3574 | maintenance |
| TIGER-AI-Lab/verl-tool VerlTool is a unified, extensible framework built on verl for training LLM agents with tool use via reinforcement learning. It decouples ac… | 61 | 1045 | active |
| metauto-ai/GPTSwarm GPTSwarm is a Python library for building LLM-based agents as computational graphs, with modules for agent graphs, memory, LLM backends, an… | 52 | 1045 | active |
| tensorflow/minigo Minigo is an open-source, minimalist implementation of the AlphaGo Zero algorithm for the game of Go, built on TensorFlow. It provides a re… | 10 | 3543 | maintenance |
| rl-tools/rl-tools RLtools is a pure C++ header-only, dependency-free deep reinforcement learning library supporting algorithms like SAC, TD3, and PPO. It com… | 66 | 1036 | active |
| silvery107/rl-mpc-locomotion A Python framework combining deep reinforcement learning with model predictive control (MPC) for quadruped robot locomotion, where a policy… | 64 | 1028 | active |
| alex-petrenko/sample-factory Sample Factory is a high-throughput Python reinforcement learning library implementing synchronous and asynchronous policy gradient algorit… | 66 | 1026 | active |
| cityflow-project/CityFlow CityFlow is an open-source microscopic traffic simulator designed as a multi-agent reinforcement learning environment for large-scale city … | 45 | 1025 | active |
| Overcooked AI Overcooked-AI is a benchmark environment for fully cooperative human-AI task performance, based on the video game Overcooked, where agents … | 25 | 1021 | active |
| AI-Optimizer AI-Optimizer is a deep reinforcement learning toolkit from TJU's DRL lab offering algorithm libraries spanning model-free, model-based, off… | 32 | 3465 | maintenance |
| ShangtongZhang/DeepRL A modularized PyTorch implementation of popular deep reinforcement learning algorithms including DQN variants, PPO, DDPG, TD3, A2C, and Opt… | 23 | 3456 | maintenance |
| pytorch/ELF ELF is an extensive, lightweight, and flexible C++/Python platform for game research, best known for ELF OpenGo, a full AlphaGoZero/AlphaZe… | 10 | 3410 | maintenance |
| MoonshotAI/checkpoint-engine Checkpoint-engine is a lightweight Python middleware for updating model weights in-place across LLM inference engines, a critical step in r… | 78 | 1005 | active |
| XYZ-AI-Lab/axrl AxisRL is an agentic reinforcement learning post-training framework for large language models, built on SGLang for high-throughput rollout … | 53 | 984 | active |
| google/brax Brax is a fast, fully differentiable rigidbody physics engine written in JAX, designed for massively parallel simulation on accelerator har… | 84 | 3247 | maintenance |
| eureka-research/Eureka Eureka is a research library from NVIDIA that uses coding LLMs like GPT-4 to automatically generate reward functions for reinforcement lear… | 28 | 3210 | maintenance |
| leggedrobotics/legged_gym A Python library providing Isaac Gym environments for training legged robots (e.g., ANYmal) to walk on rough terrain with reinforcement lea… | 39 | 3150 | maintenance |
| google-deepmind/trfl TRFL is a Python library built on TensorFlow that provides building-block loss operations (e.g., Q-learning, TD learning, distributional RL… | 32 | 3130 | maintenance |
| isaac-sim/IsaacGymEnvs A collection of example reinforcement learning environments for NVIDIA Isaac Gym, a GPU-accelerated physics simulator. It provides a Gym-st… | 10 | 2952 | maintenance |
| rail-berkeley/rlkit RLkit is a PyTorch-based reinforcement learning framework and algorithm collection from UC Berkeley RAIL. It provides reference implementat… | 32 | 2939 | maintenance |
| kzl/decision-transformer Official codebase for the Decision Transformer paper, which frames reinforcement learning as sequence modeling with transformers. It contai… | 32 | 2842 | maintenance |
| allenai/RL4LMs RL4LMs is a modular Python library from AllenAI for fine-tuning language models with reinforcement learning to align with human preferences… | 32 | 2396 | maintenance |
| hzwer/ICCV2019-LearningToPaint A PyTorch research implementation of the ICCV 2019 paper 'Learning to Paint With Model-based Deep Reinforcement Learning'. It trains agents… | 38 | 2315 | maintenance |
| astooke/rlpyt A PyTorch library providing modular, high-throughput implementations of deep reinforcement learning algorithms, covering policy gradient, d… | 32 | 2276 | maintenance |
| MineDojo MineDojo is an AI research framework built on Minecraft for training open-ended, generally capable embodied agents. It provides a simulatio… | 23 | 2260 | maintenance |
| oxwhirl/pymarl PyMARL is WhiRL's PyTorch framework for deep multi-agent reinforcement learning, implementing algorithms such as QMIX, COMA, VDN, IQL, and … | 32 | 2222 | maintenance |
| rlworkgroup/garage garage is a Python toolkit for developing and evaluating reinforcement learning algorithms, with a library of state-of-the-art RL algorithm… | 23 | 2130 | maintenance |
| sfujim/TD3 The author's official PyTorch implementation of Twin Delayed Deep Deterministic Policy Gradients (TD3), a reinforcement learning algorithm … | 32 | 2110 | maintenance |
| facebookresearch/ELF ELF is an end-to-end, lightweight and flexible C++/Python platform for game research, focused on real-time strategy games. It hosts multipl… | 10 | 2092 | maintenance |
| Yvictor/TradingGym TradingGym is an OpenAI Gym-style Python toolkit providing trading environments for training reinforcement learning agents and backtesting … | 32 | 1920 | maintenance |
| starry-sky6688/MARL-Algorithms PyTorch implementations of classic multi-agent reinforcement learning algorithms (IQL, QMIX, VDN, COMA, QTRAN, MAVEN, CommNet, DyMA-CL, G2A… | 32 | 1764 | maintenance |
| openai/Video-Pre-Training OpenAI's Video PreTraining (VPT) codebase for learning Minecraft agents by watching unlabeled online videos, including behavioral cloning a… | 37 | 1747 | maintenance |
| bwapi/bwapi BWAPI is a free and open source C++ framework for interacting with Starcraft: Broodwar, letting students, researchers, and hobbyists build … | 58 | 1702 | maintenance |
| StarCraft II API Blizzard's official StarCraft II Client API, providing protobuf protocol definitions and a reference C++ library for full external control … | 23 | 1684 | maintenance |
| Kaixhin/Rainbow A PyTorch implementation of the Rainbow deep reinforcement learning agent, combining DQN improvements like prioritized experience replay, d… | 23 | 1673 | maintenance |
| uber-research/deep-neuroevolution Distributed implementations of deep neuroevolution algorithms (ES, NS-ES, NSR-ES, DeepGA, Random Search) from Uber AI Labs research papers,… | 32 | 1667 | maintenance |
| lucidrains/PaLM-rlhf-pytorch A PyTorch library implementing Reinforcement Learning from Human Feedback (RLHF) on top of the PaLM transformer architecture, aiming to rep… | 77 | 7861 | experimental |
| karpathy/reinforcejs REINFORCEjs is a JavaScript library implementing common reinforcement learning algorithms including Dynamic Programming, Temporal Differenc… | 32 | 1488 | maintenance |
| pathak22/noreward-rl TensorFlow implementation of the ICML 2017 paper 'Curiosity-driven Exploration by Self-supervised Prediction', training RL agents with an i… | 32 | 1487 | maintenance |
| erwincoumans/motion_imitation A Python research codebase accompanying the paper 'Learning Agile Robotic Locomotion Skills by Imitating Animals'. It provides a Gym enviro… | 32 | 1470 | maintenance |
| Improbable-AI/walk-these-ways A sim-to-real reinforcement learning starter kit for the Unitree Go1 quadruped robot, implementing the Walk These Ways (MoB) locomotion con… | 32 | 1465 | maintenance |
| rail-berkeley/softlearning Softlearning is a deep reinforcement learning toolbox for training maximum entropy policies in continuous domains, and the official impleme… | 32 | 1438 | maintenance |
| facebookresearch/diplomacy_cicero Research code and model checkpoints for Cicero and Diplodocus, AI agents that play the board game Diplomacy at human level by combining lan… | 10 | 1431 | maintenance |
| Ceruleanacg/Personae Personae is a Python library implementing deep reinforcement learning (DDPG, Double DQN, Dueling DQN, Policy Gradient) and supervised learn… | 32 | 1409 | maintenance |
| opendilab/DI-star DI-star is a large-scale distributed training platform for building StarCraft II game AI, including supervised and reinforcement learning t… | 34 | 1398 | maintenance |
| TorchCraft/TorchCraft TorchCraft is a bridge library connecting the Torch machine learning framework to StarCraft: Brood War via a BWAPI module that streams game… | 10 | 1397 | maintenance |
| tinkoff-ai/CORL CORL is a research-oriented Python library providing thoroughly benchmarked single-file implementations of state-of-the-art offline and off… | 10 | 1374 | maintenance |
| Replicable-MARL/MARLlib MARLlib is a unified multi-agent reinforcement learning library built on Ray and RLlib with PyTorch, providing implementations of many MARL… | 23 | 1346 | maintenance |
| ikostrikov/pytorch-a3c A PyTorch implementation of the Asynchronous Advantage Actor Critic (A3C) reinforcement learning algorithm from the DeepMind paper 'Asynchr… | 32 | 1338 | maintenance |
| jannerm/diffuser Official code for the ICML 2022 paper 'Planning with Diffusion for Flexible Behavior Synthesis', which plans by denoising trajectories with… | 32 | 1312 | maintenance |