Mulkti-discrete now working

ml-agents/mlagents/trainers/optimizer/torch_optimizer.py

         else:
             visual_obs = []
 
-        next_obs = [torch.Tensor(next_obs[0])]
+        next_obs = np.concatenate(next_obs, axis=-1)
+        next_obs = [torch.Tensor(next_obs)]
 
         value_estimates, mean_value = self.policy.critic(vector_obs, visual_obs)
 

ml-agents/mlagents/trainers/ppo/optimizer_torch.py

         vec_obs = [torch.Tensor(vec_obs)]
         act_masks = torch.Tensor(np.array(batch["action_mask"]))
         actions = [torch.Tensor(np.array(batch["actions"]))]
+        actions = list(actions[0].permute([1, 0]))
 
         if self.policy.use_vis_obs:
             vis_obs = np.array(batch["visual_obs"])