add comments

ml-agents/mlagents/trainers/saver/saver.py

     def __init__(self):
         pass
 
+    @abc.abstractmethod
     def register(self, module: Any) -> None:
         """
         Register the modules to the Saver.

ml-agents/mlagents/trainers/torch/model_serialization.py

         dummy_masks = torch.ones(batch_dim + [sum(self.policy.actor_critic.act_size)])
         dummy_memories = torch.zeros(batch_dim + [1] + [self.policy.m_size])
 
+        # Need to pass all posslible inputs since currently keyword arguments is not
+        # supported by torch.nn.export()
+        self.dummy_input = (dummy_vec_obs, dummy_vis_obs, dummy_masks, dummy_memories)
+
+        # Input names can only contain actual input used since in torch.nn.export
+        # it maps input_names only to input nodes that exist in the graph
         self.input_names = []
         self.dynamic_axes = {"action": {0: "batch"}, "action_probs": {0: "batch"}}
         if self.policy.use_vec_obs:
             "is_continuous_control",
             "action_output_shape",
         ]
-
-        self.dummy_input = (dummy_vec_obs, dummy_vis_obs, dummy_masks, dummy_memories)
 
     def export_policy_model(self, output_filepath: str) -> None:
         """