Modular-Baselines

is a library designed for Deep reinforcement learning (DRL) research.

To keep the codebase easy to read and maintain, we’ve implemented each algorithm within a single script using the components provided in Modular-Baselines (MB). Our approach emphasizes duck-typing instead of inheritance, which helps us minimize the dependency between components and algorithms.

Algorithms

MB provides tested baseline implementation of the algorithms shown below.

Algorithm
A2C	❌	✔
PPO	✔	✔
SAC	❌	❌
TD3	❌	❌