small improvements

ml-agents/mlagents/trainers/policy/tf_policy.py

         step = self.sess.run(self.global_step)
         return step
 
-    def _set_step(self, step: int) -> int:
+    def set_step(self, step: int) -> int:
         """
         Sets current model step to step without creating additional ops.
         :param step: Step to set the current model step to.

ml-agents/mlagents/trainers/policy/torch_policy.py

         """
         return self.global_step.current_step
 
-    def _set_step(self, step: int) -> int:
+    def set_step(self, step: int) -> int:
         """
         Sets current model step to step without creating additional ops.
         :param step: Step to set the current model step to.

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

                 )
             self._check_model_version(__version__)
             if reset_global_steps:
-                self.policy._set_step(0)
+                self.policy.set_step(0)
                 logger.info(
                     "Starting training from step 0 and saving to {}.".format(
                         self.model_path

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

         for name, state_dict in saved_state_dict.items():
             self.modules[name].load_state_dict(state_dict)
         if reset_global_steps:
-            self.policy._set_step(0)
+            self.policy.set_step(0)
             logger.info(
                 "Starting training from step 0 and saving to {}.".format(
                     self.model_path

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

 class ModelSerializer:
     def __init__(self, policy):
         self.policy = policy
-        dummy_vec_obs = [torch.zeros([1] + [self.policy.vec_obs_size])]
+        # dimension for batch (and sequence_length if use recurrent)
+        dummy_dim = [1, 1] if self.policy.use_recurrent else [1]
+
+        dummy_vec_obs = [torch.zeros(dummy_dim + [self.policy.vec_obs_size])]
-            [torch.zeros([1] + list(self.policy.vis_obs_shape))]
+            [torch.zeros(dummy_dim + list(self.policy.vis_obs_shape))]
-        dummy_memories = torch.zeros([1] + [self.policy.m_size])
+        dummy_memories = torch.zeros(dummy_dim + [self.policy.m_size])
 
         self.input_names = [
             "vector_observation",
         }
         self.dummy_input = (dummy_vec_obs, dummy_vis_obs, dummy_masks, dummy_memories)
 
+    @staticmethod
     def export_policy_model(self, output_filepath: str) -> None:
         """
         Exports a Torch model for a Policy to .onnx format for Unity embedding.

ml-agents/mlagents/trainers/trainer/rl_trainer.py

         """
         pass
 
+    @staticmethod
     def create_saver(self, policy: Policy) -> BaseSaver:
         if self.framework == "torch":
             saver = TorchSaver(  # type: ignore