ApexRL Documentation¶
ApexRL is a modular reinforcement learning library designed for high-performance RL research and applications. It provides a clean, extensible architecture optimized for GPU-accelerated environments like Isaac Gym and Isaac Sim.
Key Features¶
GPU-Native Design: Optimized for CUDA-accelerated parallel environments
Modular Architecture: Easy to extend with custom algorithms, networks, and environments
Vectorized Environments: Built-in support for high-performance vectorized environments
Flexible Network Design: Custom Actor/Critic architectures with base classes
Production Ready: Comprehensive logging, checkpointing, and evaluation tools
Quick Links¶
Quick Start - Get started in 5 minutes
Your First RL Agent - Create your first RL agent
Train PPO - Train PPO step by step
Train DQN - Train DQN step by step
Train SAC - Train SAC step by step
Practice Examples - Run Genesis and Atari practice examples
API - API reference