Wen-Zhang Liu

你好，出现这个报错，是因为qmix是近期才上传到GitHub上的，尚未上传PyPI平台，所以您用"pip install xuance"指令安装的xuance版本暂时不支持。 针对这个问题，我们刚刚已经在PyPI上传了最新版的xuance，和目前GitHub上的同步了。 由于xuance会持续开发，更新，如前所述，GitHub上的时效性是最高的，我们一般在成功开发出新功能之后立刻上传至GitHub，等一些功能测试稳定后再统一上传至PyPI。 因此，如果您想保持跟进最新版的xuance，我们建议您下载GitHub上的工程文件，直接在该工程文件里运行demo也可以。

This issue can be solved by running `pip install gym[accept-rom-license]`

You can try to use "ALE/Pong-v5" as the `env_id`.

I mean, you can use "ALE/Pong-v5" instead of "Pong-v0". For example: ```python env = gym.make("ALE/Pong-v5") ``` However, this depends on the version of Gym you're using.

If you also meet this issue when installing mpi4py for xuance, you can install it via the conda command: ``` conda install mpi4py ``` After that, you can continue the...

Thank you for the advice. I have updated the ./examples folder by adding QMIX examples of SMAC environments. The examples can be found now at [https://github.com/agi-brain/xuance/tree/master/examples/qmix](https://github.com/agi-brain/xuance/tree/master/examples/qmix). You can run the...

Currently, the COMA, VDAC, and MFAC need to be improved for some bugs. We are sorry for that and will solve the bugs very soon. Thank you for the question!

Hi, the COMA algorithm works now in the newest version of XuanPolicy. Please feel free to contact me if you have any questions.

The COMA algorithm in the newest version can work now.

Hi, thank you for the advice about communication-based MARL algorithms. We will add the suggested examples you mentioned above in the future!