wenzhangliu
commented
7 months ago At present, we can see in the documentation that this project has implemented a very rich agent model algorithm and the adaptation of different types of Env. However, it seems that the summary of specific benchmark test results is relatively limited, and there are some missing results, such as Atari, MPE, MAgent, etc. None of the test results are displayed, and the MuJoCo test results are not very complete. They only include the four algorithms DDPG, TD3, A2C, and PPO, and do not distinguish between the different underlying framework implementations of PyTorch, TF, and MindSpore.
Re: Thanks for your support of XuanCe. Your advice greatly contributes to enhancing our repository. We will release more comprehensive benchmark results of XuanCe's examples across Atari, MuJoCo, SMAC, and more.
目前在文档中看到本项目实现了非常丰富的智能体模型算法，以及不同类型Env的适配，但是好像具体的benchmark试验结果汇总比较有限，存在大量的结果缺失，例如Atari、MPE、MAgent等均无试验结果展示，仅有的Mujoco试验结果也不是很完整，仅包含DDPG、TD3、A2C、PPO四个算法，而且没有区分PyTorch、TF、MindSpore不同底层框架实现。