There is a bug in def central_value_function(self, state, opponent_actions=None) in cc_mlp.py and needs to be modified.

[Maxwell-R]

If you use MultiDiscrete in Gym as the action space class, and use the mappo algorithm, an error will occur:

_File "D:\anaconda3\envs\ray\lib\site-packages\marllib\marl\models\zoo\mlp\cc_mlp.py", line 125, in central_value_function x = torch.cat([x.reshape(B, -1)] + opponent_actionsls, 1) RuntimeError: Tensors must have same number of dimensions: got 2 and 3

This is because the MultiDiscrete situation is not handled in def central_value_function(self, state, opponent_actions=None) of cc_mlp.py, resulting in different dimensions between opponent_actions_ls and [x.reshape(B, -1)].

====== 如果使用Gym的MultiDiscrete作为行动空间，同时使用MAPPO类的算法（用了cc_mlp.py的算法都会出现问题）将会出现报错： _File "D:\anaconda3\envs\ray\lib\site-packages\marllib\marl\models\zoo\mlp\cc_mlp.py", line 125, in central_value_function x = torch.cat([x.reshape(B, -1)] + opponent_actionsls, 1) RuntimeError: Tensors must have same number of dimensions: got 2 and 3 这是因为在cc_mlp.py文件里的def central_value_function(self, state, opponent_actions=None)函数中没有考虑MultiDiscrete作为动作空间的情况，从而使拼接时维数不兼容。请加上MultiDiscrete情况下opponent_actions_ls的计算结果。

[Theohhhu]

Thank you for bringing this to our attention. We are currently investigating the issue.