Tianshou: A Highly Modularized Deep Reinforcement Learning Library

Authors: Jiayi Weng, Huayu Chen, Dong Yan, Kaichao You, Alexis Duburcq, Minghao Zhang, Yi Su, Hang Su, Jun Zhu

JMLR 2022 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental To facilitate related research and prove Tianshou s reliability, we have released Tianshou s benchmark of Mu Jo Co environments, covering eight classic algorithms with state-of-the-art performance. We open-sourced Tianshou at https://github.com/thu-ml/tianshou/. Keywords: Deep Reinforcement Learning, Library, Py Torch, Modularized, Benchmark
Researcher Affiliation Collaboration Jiayi Weng EMAIL Huayu Chen EMAIL Dong Yan EMAIL Kaichao You EMAIL Alexis Duburcq EMAIL Minghao Zhang EMAIL Yi Su EMAIL Hang Su EMAIL Jun Zhu EMAIL Dept. of Comp. Sci. & Tech., BNRist Center, Institute for AI, Tsinghua-Bosch Joint ML Center, THBI Lab, Tsinghua University, Beijing, 100084, China School of Software, Tsinghua University, Beijing, 100084, China Wandercraft, 88 Rue de Rivoli, Paris, 75004, France Ant Group, 525 Almanor Ave, Sunnyvale, CA, 94085, United States of America
Pseudocode No The paper contains architectural diagrams (Figure 1 and Figure 2) but no explicit pseudocode or algorithm blocks.
Open Source Code Yes We open-sourced Tianshou at https://github.com/thu-ml/tianshou/.
Open Datasets Yes To facilitate related research and prove Tianshou s reliability, we have released Tianshou s benchmark of Mu Jo Co environments, covering eight classic algorithms with state-of-the-art performance. ... We have released Tianshou s Open AI Gym Mu Jo Co task suite benchmark, covering 8 classic algorithms and 9 environments. ... For discrete action space problems, we also provide example code and results with 7 supported algorithms in 7 Atari environments.
Dataset Splits No All experiments are done with 10 random seeds. No explicit train/test/validation splits are provided for the datasets (environments).
Hardware Specification No The paper does not provide specific details about the hardware used for running its experiments.
Software Dependencies No In this paper, we present Tianshou, a highly modularized Python library for deep reinforcement learning (DRL) that uses Py Torch as its backend. ... Tianshou has a code coverage of 94%. Every commit to Tianshou will go through unit tests on multiple platforms. We have also released a systematic benchmark of Gym s Mu Jo Co environment.
Experiment Setup No Tianshou incorporates a comprehensive set of DRL techniques for 8 benchmarked algorithms and scores 15% higher on average in terms of median performance compared with reference implementations. ... Tianshou aims to present an as-consistent-as-possible set of hyperparameters and low-level designs. While leaving the core algorithm untouched, we try to incorporate several known tricks in a specific algorithm to all similar algorithms supported by Tianshou.