52c94b7ee970da69721ca64f291e38556310482e,rllib/agents/sac/sac_torch_model.py,SACTorchModel,get_policy_output,#SACTorchModel#Any#,209

Before Change

        Returns:
            TensorType: Distribution inputs for sampling actions.
        
        return self.action_model(model_out)

    def policy_variables(self):
        Return the list of variables for the policy net.

After Change

        // Model outs may come as original Tuple observations, concat them
        // here if this is the case.
        if isinstance(self.action_model.obs_space, Box):
            if isinstance(model_out, (list, tuple)):
                model_out = torch.cat(model_out, dim=-1)
            elif isinstance(model_out, dict):
                model_out = torch.cat(list(model_out.values()), dim=-1)
        out, _ = self.action_model({"obs": model_out}, [], None)
        return out

    def policy_variables(self):

Instances