SLM-Lab is an open-source deep reinforcement learning framework in PyTorch, published on GitHub as the companion library of the book “Foundations of Deep Reinforcement Learning”. The project has gathered around 1,400 GitHub stars and is still actively maintained, with updates in 2026.
SLM-Lab is built for reproducible RL research: every algorithm, environment and hyperparameter set is declared in a single spec file, and experiments run with automatic logging, checkpointing, analysis and plots. It implements a modular family of agents โ DQN and its variants, A2C, A3C, PPO, SAC and policy-gradient methods โ that share interchangeable networks, memory buffers and exploration components, so researchers can compose new algorithms from tested parts instead of rewriting boilerplate. The lab runs single or distributed sessions, benchmarks agents across classic control, Atari and robotics-style environments, and produces the graphs and metrics needed to compare runs honestly. Because it accompanies a textbook, the codebase is deliberately readable and heavily documented, making it as useful for learning how modern deep RL fits together as for running experiments.
SLM-Lab is free and open source under the MIT license. You only need your own compute to train on.
RL researchers, students working through modern deep RL, and engineers who want a clean, hackable PyTorch lab instead of a black-box library.