[skip ci] increment steps on training

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

         self.update_buffer["advantages"].set(
             (advantages - advantages.mean()) / (advantages.std() + 1e-10)
         )
+
+        # increment steps when training instead of when generating from environment
+        self._increment_step(self.trainer_parameters["buffer_size"], self.brain_name)
         num_epoch = self.trainer_parameters["num_epoch"]
         batch_update_stats = defaultdict(list)
         for _ in range(num_epoch):

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

         :param trajectory: The Trajectory tuple containing the steps to be processed.
         """
         self._maybe_write_summary(self.get_step + len(trajectory.steps))
-        self._increment_step(len(trajectory.steps), trajectory.behavior_id)
+        # self._increment_step(len(trajectory.steps), trajectory.behavior_id)
 
     def _maybe_write_summary(self, step_after_process: int) -> None:
         """

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

 from mlagents.trainers.policy import Policy
 from mlagents.trainers.exception import UnityTrainerException
 from mlagents.trainers.behavior_id_utils import BehaviorIdentifiers
-import horovod.tensorflow as hvd
 
 
 logger = get_logger(__name__)
         stop training if it wasn't training to begin with, or if max_steps
         is reached.
         """
-        if hvd.rank() == 0:
-            return self.is_training and self.get_step <= self.get_max_steps
-        else:
-            return True
+        return self.is_training and self.get_step <= self.get_max_steps
 
     @property
     def reward_buffer(self) -> Deque[float]: