TMRL documentation

The tmrl library is a complete framework designed to help you implement deep reinforcement learning pipelines in real-world applications such as robots or videogames.

As a fun example, we readily provide a training pipeline for the TrackMania 2020 videogame.

We strongly encourage new readers to visit our GitHub as it contains a lot of information and tutorials to help you get on track :)

The documentation describes the tmrl python API and is intended for developers who want to implement their own training pipelines. We also provide an advanced tutorial for this purpose.

The three most important classes are Server, RolloutWorker and Trainer. All these classes are defined in the tmrl.networking module.

Getting started:

• Installation
• Command Line Interface

API:

• tmrl package
  • get_environment()
  • Module list
    • tmrl.networking module
      • Buffer
      • RolloutWorker
      • Server
      • Trainer
    • tmrl.actor module
      • ActorModule
      • TorchActorModule
    • tmrl.training module
      • TrainingAgent
    • tmrl.training_offline module
      • TorchTrainingOffline
      • TrainingOffline
    • tmrl.memory module
      • Memory
      • TorchMemory
    • tmrl.envs module
      • GenericGymEnv

Indices and tables

• Index

• Module Index

• Search Page