added opp, decay eps removed

Project/Assets/ML-Agents/Examples/Tennis/Prefabs/TennisArea.prefab

   m_Name: 
   m_EditorClassIdentifier: 
   m_BrainParameters:
-    vectorObservationSize: 10
+    vectorObservationSize: 14
     numStackedVectorObservations: 3
     vectorActionSize: 03000000
     vectorActionDescriptions: []
   m_Name: 
   m_EditorClassIdentifier: 
   m_BrainParameters:
-    vectorObservationSize: 10
+    vectorObservationSize: 14
     numStackedVectorObservations: 3
     vectorActionSize: 03000000
     vectorActionDescriptions: []

Project/Assets/ML-Agents/Examples/Tennis/Scripts/TennisAgent.cs

         sensor.AddObservation(m_InvertMult * m_BallRb.velocity.x / 40f);
         sensor.AddObservation(m_BallRb.velocity.y / 60f);
 
-        //sensor.AddObservation(m_InvertMult * (opponent.transform.position.x - myArea.transform.position.x) / -25f);
-        //sensor.AddObservation((opponent.transform.position.y - myArea.transform.position.y) / -7f);
-        //sensor.AddObservation(m_InvertMult * m_OpponentRb.velocity.x / 20f);
-        //sensor.AddObservation(m_OpponentRb.velocity.y / 20f);
+        sensor.AddObservation(m_InvertMult * (opponent.transform.position.x - myArea.transform.position.x) / -25f);
+        sensor.AddObservation((opponent.transform.position.y - myArea.transform.position.y) / -7f);
+        sensor.AddObservation(m_InvertMult * m_OpponentRb.velocity.x / 20f);
+        sensor.AddObservation(m_OpponentRb.velocity.y / 20f);
 
         //sensor.AddObservation(m_InvertMult * gameObject.transform.rotation.z);
         sensor.AddObservation((m_InvertMult * (gameObject.transform.rotation.eulerAngles.z - (1f - m_InvertMult) * 180f) - 35f) / 125f);

config/trainer_config.yaml

     batch_size: 2048
     buffer_size: 20480
     hidden_units: 512
-    beta: 1.0e-3
+    beta: 1.0e-2
     threaded: false
     time_horizon: 1000
     self_play:

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

         )
         advantage = tf.expand_dims(self.advantage, -1)
 
-        decay_epsilon = tf.train.polynomial_decay(
-            epsilon, self.policy.global_step, max_step, 0.1, power=1.0
-        )
+        # decay_epsilon = tf.train.polynomial_decay(
+        #    epsilon, self.policy.global_step, max_step, 0.1, power=1.0
+        # )
+        decay_epsilon = tf.Variable(epsilon)
         decay_beta = tf.Variable(beta)
 
         value_losses = []