block larger feature size; reacher fix and new reward

Project/Assets/ML-Agents/Examples/Reacher/Prefabs/Agent.prefab

   m_Name: 
   m_EditorClassIdentifier: 
   m_BrainParameters:
-    vectorObservationSize: 33
-    numStackedVectorObservations: 1
-    vectorActionSize: 04000000
-    vectorActionDescriptions: []
-    vectorActionSpaceType: 1
+    VectorObservationSize: 33
+    NumStackedVectorObservations: 1
+    VectorActionSize: 04000000
+    VectorActionDescriptions: []
+    VectorActionSpaceType: 1
   m_Model: {fileID: 11400000, guid: d7bdb6a78154f4cf99437d67e4a569a8, type: 3}
   m_InferenceDevice: 0
   m_BehaviorType: 0
+  m_ObservableAttributeHandling: 0
 --- !u!114 &114955921823023820
 MonoBehaviour:
   m_ObjectHideFlags: 0
   agentParameters:
     maxStep: 0
   hasUpgradedFromAgentParameters: 1
-  maxStep: 4000
+  MaxStep: 4000
   pendulumA: {fileID: 1644872085946016}
   pendulumB: {fileID: 1053261483945176}
   hand: {fileID: 1654288206095398}
   m_EditorClassIdentifier: 
   DecisionPeriod: 4
   TakeActionsBetweenDecisions: 1
-  offsetStep: 0
 --- !u!114 &7840105453417110232
 MonoBehaviour:
   m_ObjectHideFlags: 0
   m_Script: {fileID: 11500000, guid: 3a6da8f78a394c6ab027688eab81e04d, type: 3}
   m_Name: 
   m_EditorClassIdentifier: 
+  debugCommandLineOverride: 
 --- !u!1 &1644872085946016
 GameObject:
   m_ObjectHideFlags: 0

Project/Assets/ML-Agents/Examples/Reacher/Prefabs/NewAgent.prefab

   - component: {fileID: 33192810276213476}
   - component: {fileID: 135746602902751552}
   - component: {fileID: 23595512991530936}
-  - component: {fileID: 114928491800121992}
+  - component: {fileID: 2476992091532970082}
   m_Layer: 0
   m_Name: Goal
   m_TagString: Untagged
   m_SortingLayerID: 0
   m_SortingLayer: 0
   m_SortingOrder: 0
---- !u!114 &114928491800121992
+--- !u!114 &2476992091532970082
 MonoBehaviour:
   m_ObjectHideFlags: 0
   m_CorrespondingSourceObject: {fileID: 0}
   m_Enabled: 1
   m_EditorHideFlags: 0
-  m_Script: {fileID: 11500000, guid: a8c5005c8e6b84f1089c132cb87b44c4, type: 3}
+  m_Script: {fileID: 11500000, guid: becbcd118702d41089479cafb60baa32, type: 3}
   m_Name: 
   m_EditorClassIdentifier: 
   agent: {fileID: 1395682910799436}

Project/Assets/ML-Agents/Examples/Reacher/Scripts/NewReacherAgent.cs

         var torqueZ = Mathf.Clamp(vectorAction[1], -1f, 1f) * 150f;
         m_RbA.AddTorque(new Vector3(torqueX, 0f, torqueZ));
 
+        AddReward( - (0.05f * torqueX * torqueX + 0.05f * torqueZ * torqueZ));
+
+
+        AddReward( - (0.05f * torqueX * torqueX + 0.05f * torqueZ * torqueZ));
     }
 
     /// <summary>

Project/Assets/ML-Agents/Examples/Reacher/Scripts/ReacherAgent.cs

 
         sensor.AddObservation(goal.transform.localPosition);
         sensor.AddObservation(hand.transform.localPosition);
-
+        
         sensor.AddObservation(m_GoalSpeed);
     }
 
         var torqueZ = Mathf.Clamp(vectorAction[1], -1f, 1f) * 150f;
         m_RbA.AddTorque(new Vector3(torqueX, 0f, torqueZ));
 
+        AddReward( - (0.05f * torqueX * torqueX + 0.05f * torqueZ * torqueZ));
+
+
+        AddReward( - (0.05f * torqueX * torqueX + 0.05f * torqueZ * torqueZ));
     }
 
     /// <summary>

config/sac_transfer/PushBlock.yaml

       forward_layers: 2
       value_layers: 2
       action_layers: 2
-      feature_size: 128
+      feature_size: 256
       action_feature_size: 64
       separate_policy_train: true
       separate_policy_net: true

config/sac_transfer/PushBlockTransfer.yaml

       forward_layers: 2
       value_layers: 2
       action_layers: 2
-      feature_size: 128
+      feature_size: 256
       action_feature_size: 64
       separate_policy_train: true
       separate_policy_net: true
       train_model: false
       load_action: true
       train_action: false
-      transfer_path: "results/block/PushBlock"
+      transfer_path: "results/block-f256/PushBlock"
     network_settings:
       normalize: false
       hidden_units: 256

config/sac_transfer/Reacher.yaml

       learning_rate: 0.0003
       learning_rate_schedule: linear
       model_schedule: constant
-      batch_size: 128
+      batch_size: 256
       buffer_size: 6000000
       buffer_init_steps: 0
       tau: 0.005

config/sac_transfer/ReacherTransfer.yaml

       learning_rate: 0.0003
       learning_rate_schedule: constant
       model_schedule: constant
-      batch_size: 128
+      batch_size: 256
       buffer_size: 6000000
       buffer_init_steps: 0
       tau: 0.005
       train_model: false
       load_action: true
       train_action: false
-      transfer_path: "results/sacmod-reacher-e1/Reacher"
+      transfer_path: "results/sacmod_reacher-qr/Reacher"
     network_settings:
       normalize: true
       hidden_units: 128

Project/Assets/ML-Agents/Examples/Reacher/Scripts/NewReacherGoal.cs

+using UnityEngine;
+
+public class NewReacherGoal : MonoBehaviour
+{
+    public GameObject agent;
+    public GameObject hand;
+    public GameObject goalOn;
+
+    void OnTriggerEnter(Collider other)
+    {
+        if (other.gameObject == hand)
+        {
+            goalOn.transform.localScale = new Vector3(1f, 1f, 1f);
+        }
+    }
+
+    void OnTriggerExit(Collider other)
+    {
+        if (other.gameObject == hand)
+        {
+            goalOn.transform.localScale = new Vector3(0f, 0f, 0f);
+        }
+    }
+
+    void OnTriggerStay(Collider other)
+    {
+        if (other.gameObject == hand)
+        {
+            agent.GetComponent<NewReacherAgent>().AddReward(0.01f);
+        }
+    }
+}

Project/Assets/ML-Agents/Examples/Reacher/Scripts/NewReacherGoal.cs.meta

+fileFormatVersion: 2
+guid: becbcd118702d41089479cafb60baa32
+timeCreated: 1508689729
+licenseType: Pro
+MonoImporter:
+  serializedVersion: 2
+  defaultReferences: []
+  executionOrder: 0
+  icon: {instanceID: 0}
+  userData: 
+  assetBundleName: 
+  assetBundleVariant: