A tool for recording RL trajectories

We introduce RLDS (Reinforcement Learning Datasets), an ecosystem for recording, replaying, manipulating and sharing data in the context of Sequential Decision Making including Reinforcement Learning (RL), Learning for Demonstrations, Offline RL or Imitation Learning. RLDS enables not only reproducibility and easy generation of datasets, but also new research by providing a standard and lossless format of datasets and making it easier to run new algorithms on a wider range of datasets. By using the RLDS ecosystem, datasets can be easily shared without losing any information and researchers can apply their data processing pipelines to a collection of datasets without worrying about the underlying format. Besides, RLDS provides tools for collecting data generated by either synthetic agents or humans, as well as for inspecting and manipulating the collected data. Finally, RLDS integrates with TFDS in order to facilitate the sharing of RL datasets with the research community.

No items found.