Oct, 2022
MSRL: 数据流片段下的分布式强化学习
MSRL: Distributed Reinforcement Learning with Dataflow Fragments
Huanzhou Zhu, Bo Zhao, Gang Chen, Weifeng Chen, Yijie Chen...
TL;DRMindSpore Reinforcement Learning (MSRL) is a distributed system that supports parallel computing on GPU clusters for Reinforcement Learning algorithms without tying the distribution strategies to algorithm implementation.