8870270164e8e58f37fde60a535ce9bdbfc1cc90,rllib/agents/qmix/qmix_policy.py,QMixTorchPolicy,_unpack_observation,#QMixTorchPolicy#Any#,456

Before Change

            self.observation_space.original_space,
            tensorlib=np)
        if self.has_action_mask:
            obs = np.concatenate(
                [o["obs"] for o in unpacked],
                axis=1).reshape([len(obs_batch), self.n_agents, self.obs_size])
            action_mask = np.concatenate(
                [o["action_mask"] for o in unpacked], axis=1).reshape(
                    [len(obs_batch), self.n_agents, self.n_actions])
        else:

After Change

        if isinstance(unpacked[0], dict):
            unpacked_obs = [
                np.concatenate(tree.flatten(u["obs"]), 1) for u in unpacked
            ]
        else:
            unpacked_obs = unpacked

Instances