Fix issue exporting graph with multi-GPU (#2573)

Our multi-GPU training had a regression such that freezing the
graph was broken.  This change fixes that issue by making a few
changes:

* Removes the top level "tower" variable scope added by multi-GPU
  so that the output nodes have correct names
* Removes the use of "freeze_graph" and replaces it with our own similar 
  functionality.
* Adds the "auto reuse" to network layers which require them

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

             self.act_size[0],
             activation=None,
             kernel_initializer=LearningModel.scaled_init(0.01),
+            reuse=tf.AUTO_REUSE,
         )
 
         self.log_sigma_sq = tf.get_variable(

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

         self.devices = get_devices()
         self.towers = []
         with self.graph.as_default():
-            with tf.variable_scope(TOWER_SCOPE_NAME, reuse=tf.AUTO_REUSE):
+            with tf.variable_scope("", reuse=tf.AUTO_REUSE):
                 for device in self.devices:
                     with tf.device(device):
                         self.towers.append(
                         )
                         self.towers[-1].create_ppo_optimizer()
             self.model = self.towers[0]
-
             avg_grads = self.average_gradients([t.grads for t in self.towers])
             update_batch = self.model.optimizer.apply_gradients(avg_grads)
 

ml-agents/mlagents/trainers/tf_policy.py

 from mlagents.envs.exception import UnityException
 from mlagents.envs.policy import Policy
 from mlagents.envs.action_info import ActionInfo
-from tensorflow.python.tools import freeze_graph
+from tensorflow.python.platform import gfile
+from tensorflow.python.framework import graph_util
 from mlagents.trainers import tensorflow_to_barracuda as tf2bc
 from mlagents.envs.brain import BrainInfo
 
 
         with self.graph.as_default():
             target_nodes = ",".join(self._process_graph())
-            ckpt = tf.train.get_checkpoint_state(self.model_path)
-            freeze_graph.freeze_graph(
-                input_graph=self.model_path + "/raw_graph_def.pb",
-                input_binary=True,
-                input_checkpoint=ckpt.model_checkpoint_path,
-                output_node_names=target_nodes,
-                output_graph=(self.model_path + "/frozen_graph_def.pb"),
-                clear_devices=True,
-                initializer_nodes="",
-                input_saver="",
-                restore_op_name="save/restore_all",
-                filename_tensor_name="save/Const:0",
+            graph_def = self.graph.as_graph_def()
+            output_graph_def = graph_util.convert_variables_to_constants(
+                self.sess, graph_def, target_nodes.replace(" ", "").split(",")
-
-        tf2bc.convert(self.model_path + "/frozen_graph_def.pb", self.model_path + ".nn")
-        logger.info("Exported " + self.model_path + ".nn file")
+            frozen_graph_def_path = self.model_path + "/frozen_graph_def.pb"
+            with gfile.GFile(frozen_graph_def_path, "wb") as f:
+                f.write(output_graph_def.SerializeToString())
+            tf2bc.convert(frozen_graph_def_path, self.model_path + ".nn")
+            logger.info("Exported " + self.model_path + ".nn file")
 
     def _process_graph(self):
         """