Merge branch 'master' into self-play-mutex

.circleci/config.yml

       pip_constraints:
         type: string
         description: Constraints file that is passed to "pip install". We constraint older versions of libraries for older python runtime, in order to help ensure compatibility.
+      enforce_onnx_conversion:
+        type: integer
+        default: 0
+        description: Whether to raise an exception if ONNX models couldn't be saved.
     executor: << parameters.executor >>
     working_directory: ~/repo
 
+      TEST_ENFORCE_ONNX_CONVERSION: << parameters.enforce_onnx_conversion >>
 
     steps:
       - checkout
           pyversion: 3.7.3
           # Test python 3.7 with the newest supported versions
           pip_constraints: test_constraints_max_tf1_version.txt
+          # Make sure ONNX conversion passes here (recent version of tensorflow 1.x)
+          enforce_onnx_conversion: 1
       - build_python:
           name: python_3.7.3+tf2
           executor: python373

.pre-commit-config.yaml

         language: script
         entry: utils/validate_versions.py
         files: ".*/__init__.py"
+    -   id: validate-init-py
+        name: validate __init__.py files
+        language: script
+        entry: utils/validate_inits.py
+        types: [python]
+

.yamato/com.unity.ml-agents-test.yml

 test_editors:
   - version: 2018.4
+    # 2018.4 doesn't support code-coverage
+    coverageOptions:
+    coverageOptions: --enable-code-coverage --code-coverage-options 'generateHtmlReport;assemblyFilters:+Unity.ML-Agents'
+    coverageOptions: --enable-code-coverage --code-coverage-options 'generateHtmlReport;assemblyFilters:+Unity.ML-Agents'
 test_platforms:
   - name: win
     type: Unity::VM
     flavor: {{ platform.flavor}}
   commands:
     - npm install upm-ci-utils@stable -g --registry https://api.bintray.com/npm/unity/unity-npm
-    - upm-ci package test -u {{ editor.version }} --package-path com.unity.ml-agents
+    - upm-ci package test -u {{ editor.version }} --package-path com.unity.ml-agents {{ editor.coverageOptions }}
   artifacts:
     logs:
       paths:

DevProject/Packages/manifest.json

     "com.unity.2d.tilemap": "1.0.0",
     "com.unity.ads": "2.0.8",
     "com.unity.analytics": "3.3.5",
+    "com.unity.coding": "0.1.0-preview.13",
     "com.unity.collab-proxy": "1.2.16",
     "com.unity.ide.rider": "1.1.4",
     "com.unity.ide.vscode": "1.1.4",
     "com.unity.package-validation-suite": "0.7.15-preview",
     "com.unity.purchasing": "2.0.6",
-    "com.unity.test-framework": "1.1.9",
+    "com.unity.test-framework": "1.1.11",
-    "com.unity.timeline": "1.2.10",
+    "com.unity.timeline": "1.2.12",
     "com.unity.ugui": "1.0.0",
     "com.unity.xr.legacyinputhelpers": "1.3.8",
     "com.unity.modules.ai": "1.0.0",
     "com.unity.modules.video": "1.0.0",
     "com.unity.modules.vr": "1.0.0",
     "com.unity.modules.wind": "1.0.0",
-    "com.unity.modules.xr": "1.0.0",
-    "com.unity.coding" : "0.1.0-preview.13"
+    "com.unity.modules.xr": "1.0.0"
   },
   "registry": "https://artifactory.prd.cds.internal.unity3d.com/artifactory/api/npm/upm-candidates",
   "testables": [

DevProject/ProjectSettings/ProjectVersion.txt

-m_EditorVersion: 2019.3.0f6
-m_EditorVersionWithRevision: 2019.3.0f6 (27ab2135bccf)
+m_EditorVersion: 2019.3.3f1
+m_EditorVersionWithRevision: 2019.3.3f1 (7ceaae5f7503)

Project/Assets/ML-Agents/Examples/3DBall/Scenes/3DBall.unity

   m_ReflectionIntensity: 1
   m_CustomReflection: {fileID: 0}
   m_Sun: {fileID: 0}
-  m_IndirectSpecularColor: {r: 0.4497121, g: 0.49977785, b: 0.57563704, a: 1}
+  m_IndirectSpecularColor: {r: 0.44971162, g: 0.49977726, b: 0.5756362, a: 1}
   m_UseRadianceAmbientProbe: 0
 --- !u!157 &3
 LightmapSettings:
   m_Component:
   - component: {fileID: 807556627}
   - component: {fileID: 807556626}
-  - component: {fileID: 807556625}
   - component: {fileID: 807556624}
   - component: {fileID: 807556623}
   m_Layer: 0
   m_PrefabAsset: {fileID: 0}
   m_GameObject: {fileID: 807556622}
   m_Enabled: 1
---- !u!92 &807556625
-Behaviour:
-  m_ObjectHideFlags: 0
-  m_CorrespondingSourceObject: {fileID: 0}
-  m_PrefabInstance: {fileID: 0}
-  m_PrefabAsset: {fileID: 0}
-  m_GameObject: {fileID: 807556622}
-  m_Enabled: 1
 --- !u!20 &807556626
 Camera:
   m_ObjectHideFlags: 0
   m_Name: 
   m_EditorClassIdentifier: 
   gravityMultiplier: 1
-  monitorVerticalOffset: 0
   fixedDeltaTime: 0.02
   maximumDeltaTime: 0.33333334
   solverIterations: 6

Project/Assets/ML-Agents/Examples/3DBall/Scenes/3DBallHard.unity

   m_ReflectionIntensity: 1
   m_CustomReflection: {fileID: 0}
   m_Sun: {fileID: 0}
-  m_IndirectSpecularColor: {r: 0.4497121, g: 0.49977785, b: 0.57563704, a: 1}
+  m_IndirectSpecularColor: {r: 0.44971162, g: 0.49977726, b: 0.5756362, a: 1}
   m_UseRadianceAmbientProbe: 0
 --- !u!157 &3
 LightmapSettings:
   m_Component:
   - component: {fileID: 807556627}
   - component: {fileID: 807556626}
-  - component: {fileID: 807556625}
   - component: {fileID: 807556624}
   - component: {fileID: 807556623}
   m_Layer: 0
   m_GameObject: {fileID: 807556622}
   m_Enabled: 1
 --- !u!124 &807556624
-Behaviour:
-  m_ObjectHideFlags: 0
-  m_CorrespondingSourceObject: {fileID: 0}
-  m_PrefabInstance: {fileID: 0}
-  m_PrefabAsset: {fileID: 0}
-  m_GameObject: {fileID: 807556622}
-  m_Enabled: 1
---- !u!92 &807556625
 Behaviour:
   m_ObjectHideFlags: 0
   m_CorrespondingSourceObject: {fileID: 0}
   m_Name: 
   m_EditorClassIdentifier: 
   gravityMultiplier: 1
-  monitorVerticalOffset: 0
   fixedDeltaTime: 0.02
   maximumDeltaTime: 0.33333334
   solverIterations: 6

Project/Assets/ML-Agents/Examples/3DBall/Scripts/Ball3DAgent.cs

 using UnityEngine;
 using MLAgents;
 using MLAgents.Sensors;
+using MLAgents.SideChannels;
 
 public class Ball3DAgent : Agent
 {
-    IFloatProperties m_ResetParams;
+    FloatPropertiesChannel m_ResetParams;
-    public override void InitializeAgent()
+    public override void Initialize()
     {
         m_BallRb = ball.GetComponent<Rigidbody>();
         m_ResetParams = Academy.Instance.FloatProperties;
         sensor.AddObservation(m_BallRb.velocity);
     }
 
-    public override void AgentAction(float[] vectorAction)
+    public override void OnActionReceived(float[] vectorAction)
     {
         var actionZ = 2f * Mathf.Clamp(vectorAction[0], -1f, 1f);
         var actionX = 2f * Mathf.Clamp(vectorAction[1], -1f, 1f);
             Mathf.Abs(ball.transform.position.z - gameObject.transform.position.z) > 3f)
         {
             SetReward(-1f);
-            Done();
+            EndEpisode();
         }
         else
         {
 
-    public override void AgentReset()
+    public override void OnEpisodeBegin()
     {
         gameObject.transform.rotation = new Quaternion(0f, 0f, 0f, 0f);
         gameObject.transform.Rotate(new Vector3(1, 0, 0), Random.Range(-10f, 10f));

Project/Assets/ML-Agents/Examples/3DBall/Scripts/Ball3DHardAgent.cs

 using UnityEngine;
 using MLAgents;
 using MLAgents.Sensors;
+using MLAgents.SideChannels;
 
 public class Ball3DHardAgent : Agent
 {
-    IFloatProperties m_ResetParams;
+    FloatPropertiesChannel m_ResetParams;
-    public override void InitializeAgent()
+    public override void Initialize()
     {
         m_BallRb = ball.GetComponent<Rigidbody>();
         m_ResetParams = Academy.Instance.FloatProperties;
         sensor.AddObservation((ball.transform.position - gameObject.transform.position));
     }
 
-    public override void AgentAction(float[] vectorAction)
+    public override void OnActionReceived(float[] vectorAction)
     {
         var actionZ = 2f * Mathf.Clamp(vectorAction[0], -1f, 1f);
         var actionX = 2f * Mathf.Clamp(vectorAction[1], -1f, 1f);
             Mathf.Abs(ball.transform.position.z - gameObject.transform.position.z) > 3f)
         {
             SetReward(-1f);
-            Done();
+            EndEpisode();
         }
         else
         {
 
-    public override void AgentReset()
+    public override void OnEpisodeBegin()
     {
         gameObject.transform.rotation = new Quaternion(0f, 0f, 0f, 0f);
         gameObject.transform.Rotate(new Vector3(1, 0, 0), Random.Range(-10f, 10f));

Project/Assets/ML-Agents/Examples/Basic/Scenes/Basic.unity

       objectReference: {fileID: 0}
     m_RemovedComponents: []
   m_SourcePrefab: {fileID: 100100000, guid: 5889392e3f05b448a8a06c5def6c2dec, type: 3}
+--- !u!1 &1335907378 stripped
+GameObject:
+  m_CorrespondingSourceObject: {fileID: 1263463520136984, guid: c5eb289873aca4f5a8cc59c7464ab7c1,
+    type: 3}
+  m_PrefabInstance: {fileID: 1783603361}
+  m_PrefabAsset: {fileID: 0}
+--- !u!114 &1335907380
+MonoBehaviour:
+  m_ObjectHideFlags: 0
+  m_CorrespondingSourceObject: {fileID: 0}
+  m_PrefabInstance: {fileID: 0}
+  m_PrefabAsset: {fileID: 0}
+  m_GameObject: {fileID: 1335907378}
+  m_Enabled: 1
+  m_EditorHideFlags: 0
+  m_Script: {fileID: 11500000, guid: 88b6042bc9a5d4aa58d931eae49442e5, type: 3}
+  m_Name: 
+  m_EditorClassIdentifier: 
+  agentParameters:
+    maxStep: 0
+  hasUpgradedFromAgentParameters: 1
+  maxStep: 0
+--- !u!114 &1335907381 stripped
+MonoBehaviour:
+  m_CorrespondingSourceObject: {fileID: 114827551040495112, guid: c5eb289873aca4f5a8cc59c7464ab7c1,
+    type: 3}
+  m_PrefabInstance: {fileID: 1783603361}
+  m_PrefabAsset: {fileID: 0}
+  m_GameObject: {fileID: 1335907378}
+  m_Enabled: 1
+  m_EditorHideFlags: 0
+  m_Script: {fileID: 11500000, guid: 624480a72e46148118ab2e2d89b537de, type: 3}
+  m_Name: 
+  m_EditorClassIdentifier: 
+--- !u!114 &1335907384
+MonoBehaviour:
+  m_ObjectHideFlags: 0
+  m_CorrespondingSourceObject: {fileID: 0}
+  m_PrefabInstance: {fileID: 0}
+  m_PrefabAsset: {fileID: 0}
+  m_GameObject: {fileID: 1335907378}
+  m_Enabled: 1
+  m_EditorHideFlags: 0
+  m_Script: {fileID: 11500000, guid: 6ee410d6d45349218d5e69bb2a347c63, type: 3}
+  m_Name: 
+  m_EditorClassIdentifier: 
+  basicController: {fileID: 1335907381}
 --- !u!1001 &1502457254
 PrefabInstance:
   m_ObjectHideFlags: 0
   m_Component:
   - component: {fileID: 1715640925}
   - component: {fileID: 1715640924}
-  - component: {fileID: 1715640923}
   - component: {fileID: 1715640922}
   - component: {fileID: 1715640921}
   m_Layer: 0
   m_GameObject: {fileID: 1715640920}
   m_Enabled: 1
 --- !u!124 &1715640922
-Behaviour:
-  m_ObjectHideFlags: 0
-  m_CorrespondingSourceObject: {fileID: 0}
-  m_PrefabInstance: {fileID: 0}
-  m_PrefabAsset: {fileID: 0}
-  m_GameObject: {fileID: 1715640920}
-  m_Enabled: 1
---- !u!92 &1715640923
 Behaviour:
   m_ObjectHideFlags: 0
   m_CorrespondingSourceObject: {fileID: 0}
       propertyPath: m_LocalScale.z
       value: 6.270299
       objectReference: {fileID: 0}
+    - target: {fileID: 114502619508238574, guid: c5eb289873aca4f5a8cc59c7464ab7c1,
+        type: 3}
+      propertyPath: m_BrainParameters.vectorObservationSize
+      value: 0
+      objectReference: {fileID: 0}
     m_RemovedComponents: []
   m_SourcePrefab: {fileID: 100100000, guid: c5eb289873aca4f5a8cc59c7464ab7c1, type: 3}
 --- !u!1 &1889211226
   m_Name: 
   m_EditorClassIdentifier: 
   gravityMultiplier: 1
-  monitorVerticalOffset: 0
   fixedDeltaTime: 0.02
   maximumDeltaTime: 0.33333334
   solverIterations: 6

Project/Assets/ML-Agents/Examples/Bouncer/Scripts/BouncerAgent.cs

 using UnityEngine;
 using MLAgents;
 using MLAgents.Sensors;
+using MLAgents.SideChannels;
 
 public class BouncerAgent : Agent
 {
     int m_NumberJumps = 20;
     int m_JumpLeft = 20;
 
-    IFloatProperties m_ResetParams;
+    FloatPropertiesChannel m_ResetParams;
-    public override void InitializeAgent()
+    public override void Initialize()
     {
         m_Rb = gameObject.GetComponent<Rigidbody>();
         m_LookDir = Vector3.zero;
         sensor.AddObservation(target.transform.localPosition);
     }
 
-    public override void AgentAction(float[] vectorAction)
+    public override void OnActionReceived(float[] vectorAction)
     {
         for (var i = 0; i < vectorAction.Length; i++)
         {
         m_LookDir = new Vector3(x, y, z);
     }
 
-    public override void AgentReset()
+    public override void OnEpisodeBegin()
     {
         gameObject.transform.localPosition = new Vector3(
             (1 - 2 * Random.value) * 5, 2, (1 - 2 * Random.value) * 5);
         if (gameObject.transform.position.y < -1)
         {
             AddReward(-1);
-            Done();
+            EndEpisode();
             return;
         }
 
             AddReward(-1);
-            Done();
+            EndEpisode();
-            Done();
+            EndEpisode();
         }
     }
 

Project/Assets/ML-Agents/Examples/Crawler/Scenes/CrawlerDynamicTarget.unity

   m_Component:
   - component: {fileID: 1392866532}
   - component: {fileID: 1392866531}
-  - component: {fileID: 1392866530}
   - component: {fileID: 1392866529}
   - component: {fileID: 1392866528}
   - component: {fileID: 1392866533}
   m_GameObject: {fileID: 1392866527}
   m_Enabled: 1
 --- !u!124 &1392866529
-Behaviour:
-  m_ObjectHideFlags: 0
-  m_CorrespondingSourceObject: {fileID: 0}
-  m_PrefabInstance: {fileID: 0}
-  m_PrefabAsset: {fileID: 0}
-  m_GameObject: {fileID: 1392866527}
-  m_Enabled: 1
---- !u!92 &1392866530
 Behaviour:
   m_ObjectHideFlags: 0
   m_CorrespondingSourceObject: {fileID: 0}
   m_Name: 
   m_EditorClassIdentifier: 
   gravityMultiplier: 1
-  monitorVerticalOffset: 1
   fixedDeltaTime: 0.01333
   maximumDeltaTime: 0.15
   solverIterations: 12

Project/Assets/ML-Agents/Examples/Crawler/Scripts/CrawlerAgent.cs

     Quaternion m_LookRotation;
     Matrix4x4 m_TargetDirMatrix;
 
-    public override void InitializeAgent()
+    public override void Initialize()
     {
         m_JdController = GetComponent<JointDriveController>();
         m_DirToTarget = target.position - body.position;
         target.position = newTargetPos + ground.position;
     }
 
-    public override void AgentAction(float[] vectorAction)
+    public override void OnActionReceived(float[] vectorAction)
     {
         // The dictionary with all the body parts in it are in the jdController
         var bpDict = m_JdController.bodyPartsDict;
     /// <summary>
     /// Loop over body parts and reset them to initial conditions.
     /// </summary>
-    public override void AgentReset()
+    public override void OnEpisodeBegin()
     {
         if (m_DirToTarget != Vector3.zero)
         {

Project/Assets/ML-Agents/Examples/FoodCollector/Scripts/FoodCollectorAgent.cs

     public bool useVectorObs;
 
 
-    public override void InitializeAgent()
+    public override void Initialize()
-        base.InitializeAgent();
         m_AgentRb = GetComponent<Rigidbody>();
         m_MyArea = area.GetComponent<FoodCollectorArea>();
         m_FoodCollecterSettings = FindObjectOfType<FoodCollectorSettings>();
         gameObject.GetComponentInChildren<Renderer>().material = normalMaterial;
     }
 
-    public override void AgentAction(float[] vectorAction)
+    public override void OnActionReceived(float[] vectorAction)
     {
         MoveAgent(vectorAction);
     }
         return action;
     }
 
-    public override void AgentReset()
+    public override void OnEpisodeBegin()
     {
         Unfreeze();
         Unpoison();

Project/Assets/ML-Agents/Examples/GridWorld/Scenes/GridWorld.unity

   m_ReflectionIntensity: 1
   m_CustomReflection: {fileID: 0}
   m_Sun: {fileID: 0}
-  m_IndirectSpecularColor: {r: 0.4497121, g: 0.49977785, b: 0.57563704, a: 1}
+  m_IndirectSpecularColor: {r: 0.44971162, g: 0.49977726, b: 0.5756362, a: 1}
   m_UseRadianceAmbientProbe: 0
 --- !u!157 &3
 LightmapSettings:
   m_Component:
   - component: {fileID: 99095116}
   - component: {fileID: 99095115}
-  - component: {fileID: 99095114}
   - component: {fileID: 99095113}
   m_Layer: 0
   m_Name: Main Camera
   m_PrefabAsset: {fileID: 0}
   m_GameObject: {fileID: 99095112}
   m_Enabled: 1
---- !u!92 &99095114
-Behaviour:
-  m_ObjectHideFlags: 0
-  m_CorrespondingSourceObject: {fileID: 0}
-  m_PrefabInstance: {fileID: 0}
-  m_PrefabAsset: {fileID: 0}
-  m_GameObject: {fileID: 99095112}
-  m_Enabled: 1
 --- !u!20 &99095115
 Camera:
   m_ObjectHideFlags: 0
   m_EditorClassIdentifier: 
   agentParameters:
     maxStep: 100
-    resetOnDone: 1
-    onDemandDecision: 1
-    numberOfActionsBetweenDecisions: 1
+  hasUpgradedFromAgentParameters: 1
+  maxStep: 100
   area: {fileID: 1795599557}
   timeBetweenDecisionsAtInference: 0.15
   renderCamera: {fileID: 797520692}
   m_InferenceDevice: 0
   m_BehaviorType: 0
   m_BehaviorName: GridWorld
-  m_TeamID: 0
-  m_useChildSensors: 1
+  TeamId: 0
+  m_UseChildSensors: 1
 --- !u!114 &125487791
 MonoBehaviour:
   m_ObjectHideFlags: 0
   m_Script: {fileID: 11500000, guid: 132e1194facb64429b007ea1edf562d0, type: 3}
   m_Name: 
   m_EditorClassIdentifier: 
-  renderTexture: {fileID: 8400000, guid: 114608d5384404f89bff4b6f88432958, type: 2}
-  sensorName: RenderTextureSensor
-  grayscale: 0
-  compression: 1
+  m_RenderTexture: {fileID: 8400000, guid: 114608d5384404f89bff4b6f88432958, type: 2}
+  m_SensorName: RenderTextureSensor
+  m_Grayscale: 0
+  m_Compression: 1
 --- !u!1 &260425459
 GameObject:
   m_ObjectHideFlags: 0

Project/Assets/ML-Agents/Examples/GridWorld/Scripts/GridAgent.cs

     const int k_Left = 3;
     const int k_Right = 4;
 
-    public override void InitializeAgent()
-    {
-    }
-
     public override void CollectDiscreteActionMasks(DiscreteActionMasker actionMasker)
     {
         // Mask the necessary actions if selected by the user.
     }
 
     // to be implemented by the developer
-    public override void AgentAction(float[] vectorAction)
+    public override void OnActionReceived(float[] vectorAction)
     {
         AddReward(-0.01f);
         var action = Mathf.FloorToInt(vectorAction[0]);
             if (hit.Where(col => col.gameObject.CompareTag("goal")).ToArray().Length == 1)
             {
                 SetReward(1f);
-                Done();
+                EndEpisode();
-                Done();
+                EndEpisode();
             }
         }
     }
     }
 
     // to be implemented by the developer
-    public override void AgentReset()
+    public override void OnEpisodeBegin()
     {
         area.AreaReset();
     }

Project/Assets/ML-Agents/Examples/GridWorld/Scripts/GridArea.cs

 using UnityEngine;
 using System.Linq;
 using MLAgents;
+using MLAgents.SideChannels;
 
 
 public class GridArea : MonoBehaviour
 
     public GameObject trueAgent;
 
-    IFloatProperties m_ResetParameters;
+    FloatPropertiesChannel m_ResetParameters;
 
     Camera m_AgentCam;
 

Project/Assets/ML-Agents/Examples/Hallway/Scripts/HallwayAgent.cs

     HallwaySettings m_HallwaySettings;
     int m_Selection;
 
-    public override void InitializeAgent()
+    public override void Initialize()
-        base.InitializeAgent();
         m_HallwaySettings = FindObjectOfType<HallwaySettings>();
         m_AgentRb = GetComponent<Rigidbody>();
         m_GroundRenderer = ground.GetComponent<Renderer>();
         m_AgentRb.AddForce(dirToGo * m_HallwaySettings.agentRunSpeed, ForceMode.VelocityChange);
     }
 
-    public override void AgentAction(float[] vectorAction)
+    public override void OnActionReceived(float[] vectorAction)
     {
         AddReward(-1f / maxStep);
         MoveAgent(vectorAction);
                 SetReward(-0.1f);
                 StartCoroutine(GoalScoredSwapGroundMaterial(m_HallwaySettings.failMaterial, 0.5f));
             }
-            Done();
+            EndEpisode();
         }
     }
 
         return new float[] { 0 };
     }
 
-    public override void AgentReset()
+    public override void OnEpisodeBegin()
     {
         var agentOffset = -15f;
         var blockOffset = 0f;

Project/Assets/ML-Agents/Examples/PushBlock/Scripts/PushAgentBasic.cs

         m_PushBlockSettings = FindObjectOfType<PushBlockSettings>();
     }
 
-    public override void InitializeAgent()
+    public override void Initialize()
-        base.InitializeAgent();
         goalDetect = block.GetComponent<GoalDetect>();
         goalDetect.agent = this;
 
         AddReward(5f);
 
         // By marking an agent as done AgentReset() will be called automatically.
-        Done();
+        EndEpisode();
 
         // Swap ground material for a bit to indicate we scored.
         StartCoroutine(GoalScoredSwapGroundMaterial(m_PushBlockSettings.goalScoredMaterial, 0.5f));
     /// <summary>
     /// Called every step of the engine. Here the agent takes an action.
     /// </summary>
-    public override void AgentAction(float[] vectorAction)
+    public override void OnActionReceived(float[] vectorAction)
     {
         // Move the agent using the action.
         MoveAgent(vectorAction);
     /// In the editor, if "Reset On Done" is checked then AgentReset() will be
     /// called automatically anytime we mark done = true in an agent script.
     /// </summary>
-    public override void AgentReset()
+    public override void OnEpisodeBegin()
     {
         var rotation = Random.Range(0, 4);
         var rotationAngle = rotation * 90f;

Project/Assets/ML-Agents/Examples/Pyramids/Scripts/PyramidAgent.cs

     public GameObject areaSwitch;
     public bool useVectorObs;
 
-    public override void InitializeAgent()
+    public override void Initialize()
-        base.InitializeAgent();
         m_AgentRb = GetComponent<Rigidbody>();
         m_MyArea = area.GetComponent<PyramidArea>();
         m_SwitchLogic = areaSwitch.GetComponent<PyramidSwitch>();
         m_AgentRb.AddForce(dirToGo * 2f, ForceMode.VelocityChange);
     }
 
-    public override void AgentAction(float[] vectorAction)
+    public override void OnActionReceived(float[] vectorAction)
     {
         AddReward(-1f / maxStep);
         MoveAgent(vectorAction);
         return new float[] { 0 };
     }
 
-    public override void AgentReset()
+    public override void OnEpisodeBegin()
     {
         var enumerable = Enumerable.Range(0, 9).OrderBy(x => Guid.NewGuid()).Take(9);
         var items = enumerable.ToArray();
         if (collision.gameObject.CompareTag("goal"))
         {
             SetReward(2f);
-            Done();
+            EndEpisode();
         }
     }
 }

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

     /// Collect the rigidbodies of the reacher in order to resue them for
     /// observations and actions.
     /// </summary>
-    public override void InitializeAgent()
+    public override void Initialize()
     {
         m_RbA = pendulumA.GetComponent<Rigidbody>();
         m_RbB = pendulumB.GetComponent<Rigidbody>();
     /// <summary>
     /// The agent's four actions correspond to torques on each of the two joints.
     /// </summary>
-    public override void AgentAction(float[] vectorAction)
+    public override void OnActionReceived(float[] vectorAction)
     {
         m_GoalDegree += m_GoalSpeed;
         UpdateGoalPosition();
     /// <summary>
     /// Resets the position and velocity of the agent and the goal.
     /// </summary>
-    public override void AgentReset()
+    public override void OnEpisodeBegin()
     {
         pendulumA.transform.position = new Vector3(0f, -4f, 0f) + transform.position;
         pendulumA.transform.rotation = Quaternion.Euler(180f, 0f, 0f);

Project/Assets/ML-Agents/Examples/SharedAssets/Scripts/GroundContact.cs

 
                 if (agentDoneOnGroundContact)
                 {
-                    agent.Done();
+                    agent.EndEpisode();
                 }
             }
         }

Project/Assets/ML-Agents/Examples/SharedAssets/Scripts/ModelOverrider.cs

 using Barracuda;
 using System.IO;
 using MLAgents;
+using MLAgents.Policies;
 
 namespace MLAgentsExamples
 {
         {
             m_Agent.LazyInitialize();
             var bp = m_Agent.GetComponent<BehaviorParameters>();
+            var name = bp.behaviorName;
-            var nnModel = GetModelForBehaviorName(bp.behaviorName);
-            Debug.Log($"Overriding behavior {bp.behaviorName} for agent with model {nnModel?.name}");
+            var nnModel = GetModelForBehaviorName(name);
+            Debug.Log($"Overriding behavior {name} for agent with model {nnModel?.name}");
-            m_Agent.GiveModel($"Override_{bp.behaviorName}", nnModel);
+            m_Agent.SetModel($"Override_{name}", nnModel);
 
         }
     }

Project/Assets/ML-Agents/Examples/Soccer/Scripts/AgentSoccer.cs

 using System;
 using UnityEngine;
 using MLAgents;
+using MLAgents.Policies;
 
 public class AgentSoccer : Agent
 {
     BehaviorParameters m_BehaviorParameters;
     Vector3 m_Transform;
 
-    public override void InitializeAgent()
+    public override void Initialize()
-        base.InitializeAgent();
-        if (m_BehaviorParameters.m_TeamID == (int)Team.Blue)
+        if (m_BehaviorParameters.TeamId == (int)Team.Blue)
         {
             team = Team.Blue;
             m_Transform = new Vector3(transform.position.x - 4f, .5f, transform.position.z);
             ForceMode.VelocityChange);
     }
 
-    public override void AgentAction(float[] vectorAction)
+    public override void OnActionReceived(float[] vectorAction)
     {
         // Existential penalty for strikers.
         AddReward(-1f / 3000f);
         }
     }
 
-    public override void AgentReset()
+    public override void OnEpisodeBegin()
     {
         if (team == Team.Purple)
         {

Project/Assets/ML-Agents/Examples/Soccer/Scripts/SoccerFieldArea.cs

             {
                 ps.agentScript.AddReward(-1);
             }
-            ps.agentScript.Done();  //all agents need to be reset
+            ps.agentScript.EndEpisode();  //all agents need to be reset
 
             if (goalTextUI)
             {

Project/Assets/ML-Agents/Examples/Startup/Scripts/Startup.cs

     internal class Startup : MonoBehaviour
     {
         const string k_SceneVariableName = "SCENE_NAME";
+        private const string k_SceneCommandLineFlag = "--mlagents-scene-name";
-            var sceneName = Environment.GetEnvironmentVariable(k_SceneVariableName);
+            var sceneName = "";
+            
+            // Check for the CLI '--scene-name' flag.  This will be used if
+            // no scene environment variable is found.
+            var args = Environment.GetCommandLineArgs();
+            Console.WriteLine("Command line arguments passed: " + String.Join(" ", args));
+            for (int i = 0; i < args.Length; i++) {
+                if (args [i] == k_SceneCommandLineFlag && i < args.Length - 1) {
+                    sceneName = args[i + 1];
+                }
+            }
+
+            var sceneEnvironmentVariable = Environment.GetEnvironmentVariable(k_SceneVariableName);
+            if (!string.IsNullOrEmpty(sceneEnvironmentVariable))
+            {
+                sceneName = sceneEnvironmentVariable;
+            }
+            
             SwitchScene(sceneName);
         }
 
             {
-                throw new ArgumentException(
-                    $"You didn't specified the {k_SceneVariableName} environment variable");
+                Console.WriteLine(
+                    $"You didn't specify the {k_SceneVariableName} environment variable or the {k_SceneCommandLineFlag} command line argument."
+                );
+                Application.Quit(22);
+                return;
-                throw new ArgumentException(
-                    $"The scene {sceneName} doesn't exist within your build. ");
+                Console.WriteLine(
+                    $"The scene {sceneName} doesn't exist within your build."
+                );
+                Application.Quit(22);
+                return;
             }
             SceneManager.LoadSceneAsync(sceneName);
         }

Project/Assets/ML-Agents/Examples/Template/Scene.unity

 --- !u!104 &2
 RenderSettings:
   m_ObjectHideFlags: 0
-  serializedVersion: 8
+  serializedVersion: 9
   m_Fog: 0
   m_FogColor: {r: 0.5, g: 0.5, b: 0.5, a: 1}
   m_FogMode: 3
   m_CustomReflection: {fileID: 0}
   m_Sun: {fileID: 0}
   m_IndirectSpecularColor: {r: 0, g: 0, b: 0, a: 1}
+  m_UseRadianceAmbientProbe: 0
 --- !u!157 &3
 LightmapSettings:
   m_ObjectHideFlags: 0
     m_BounceScale: 1
     m_IndirectOutputScale: 1
     m_AlbedoBoost: 1
-    m_TemporalCoherenceThreshold: 1
-    serializedVersion: 9
+    serializedVersion: 10
-    m_TextureWidth: 1024
-    m_TextureHeight: 1024
+    m_AtlasSize: 1024
     m_AO: 0
     m_AOMaxDistance: 1
     m_CompAOExponent: 1
 --- !u!1 &762086410
 GameObject:
   m_ObjectHideFlags: 0
-  m_PrefabParentObject: {fileID: 0}
-  m_PrefabInternal: {fileID: 0}
-  serializedVersion: 5
+  m_CorrespondingSourceObject: {fileID: 0}
+  m_PrefabInstance: {fileID: 0}
+  m_PrefabAsset: {fileID: 0}
+  serializedVersion: 6
   m_Component:
   - component: {fileID: 762086412}
   - component: {fileID: 762086411}
 --- !u!108 &762086411
 Light:
   m_ObjectHideFlags: 0
-  m_PrefabParentObject: {fileID: 0}
-  m_PrefabInternal: {fileID: 0}
+  m_CorrespondingSourceObject: {fileID: 0}
+  m_PrefabInstance: {fileID: 0}
+  m_PrefabAsset: {fileID: 0}
   m_GameObject: {fileID: 762086410}
   m_Enabled: 1
   serializedVersion: 8
     serializedVersion: 2
     m_Bits: 4294967295
   m_Lightmapping: 4
+  m_LightShadowCasterMode: 0
   m_AreaSize: {x: 1, y: 1}
   m_BounceIntensity: 1
   m_ColorTemperature: 6570
 --- !u!4 &762086412
 Transform:
   m_ObjectHideFlags: 0
-  m_PrefabParentObject: {fileID: 0}
-  m_PrefabInternal: {fileID: 0}
+  m_CorrespondingSourceObject: {fileID: 0}
+  m_PrefabInstance: {fileID: 0}
+  m_PrefabAsset: {fileID: 0}
   m_GameObject: {fileID: 762086410}
   m_LocalRotation: {x: 0.40821788, y: -0.23456968, z: 0.10938163, w: 0.8754261}
   m_LocalPosition: {x: 0, y: 3, z: 0}
 --- !u!1 &1223085755
 GameObject:
   m_ObjectHideFlags: 0
-  m_PrefabParentObject: {fileID: 0}
-  m_PrefabInternal: {fileID: 0}
-  serializedVersion: 5
+  m_CorrespondingSourceObject: {fileID: 0}
+  m_PrefabInstance: {fileID: 0}
+  m_PrefabAsset: {fileID: 0}
+  serializedVersion: 6
   m_Component:
   - component: {fileID: 1223085757}
   - component: {fileID: 1223085756}
 --- !u!114 &1223085756
 MonoBehaviour:
   m_ObjectHideFlags: 0
-  m_PrefabParentObject: {fileID: 0}
-  m_PrefabInternal: {fileID: 0}
+  m_CorrespondingSourceObject: {fileID: 0}
+  m_PrefabInstance: {fileID: 0}
+  m_PrefabAsset: {fileID: 0}
   m_GameObject: {fileID: 1223085755}
   m_Enabled: 1
   m_EditorHideFlags: 0
-  brain: {fileID: 0}
-    agentCameras: []
-    resetOnDone: 1
-    onDemandDecision: 0
-    numberOfActionsBetweenDecisions: 1
+  hasUpgradedFromAgentParameters: 1
+  maxStep: 0
-  m_PrefabParentObject: {fileID: 0}
-  m_PrefabInternal: {fileID: 0}
+  m_CorrespondingSourceObject: {fileID: 0}
+  m_PrefabInstance: {fileID: 0}
+  m_PrefabAsset: {fileID: 0}
   m_GameObject: {fileID: 1223085755}
   m_LocalRotation: {x: 0, y: 0, z: 0, w: 1}
   m_LocalPosition: {x: 0.71938086, y: 0.27357092, z: 4.1970553}
 --- !u!1 &1574236047
 GameObject:
   m_ObjectHideFlags: 0
-  m_PrefabParentObject: {fileID: 0}
-  m_PrefabInternal: {fileID: 0}
-  serializedVersion: 5
+  m_CorrespondingSourceObject: {fileID: 0}
+  m_PrefabInstance: {fileID: 0}
+  m_PrefabAsset: {fileID: 0}
+  serializedVersion: 6
-  - component: {fileID: 1574236048}
   m_Layer: 0
   m_Name: Academy
   m_TagString: Untagged
   m_IsActive: 1
---- !u!114 &1574236048
-MonoBehaviour:
-  m_ObjectHideFlags: 0
-  m_PrefabParentObject: {fileID: 0}
-  m_PrefabInternal: {fileID: 0}
-  m_GameObject: {fileID: 1574236047}
-  m_Enabled: 1
-  m_EditorHideFlags: 0
-  m_Script: {fileID: 11500000, guid: 9af83cd96d4bc4088a966af174446d1b, type: 3}
-  m_Name: 
-  m_EditorClassIdentifier: 
-  broadcastHub:
-    broadcastingBrains: []
-    _brainsToControl: []
-  maxSteps: 0
-  trainingConfiguration:
-    width: 80
-    height: 80
-    qualityLevel: 0
-    timeScale: 100
-    targetFrameRate: 60
-  inferenceConfiguration:
-    width: 1024
-    height: 720
-    qualityLevel: 1
-    timeScale: 1
-    targetFrameRate: 60
-  resetParameters:
-    resetParameters: []
-  m_PrefabParentObject: {fileID: 0}
-  m_PrefabInternal: {fileID: 0}
+  m_CorrespondingSourceObject: {fileID: 0}
+  m_PrefabInstance: {fileID: 0}
+  m_PrefabAsset: {fileID: 0}
   m_GameObject: {fileID: 1574236047}
   m_LocalRotation: {x: 0, y: 0, z: 0, w: 1}
   m_LocalPosition: {x: 0.71938086, y: 0.27357092, z: 4.1970553}
 --- !u!1 &1715640920
 GameObject:
   m_ObjectHideFlags: 0
-  m_PrefabParentObject: {fileID: 0}
-  m_PrefabInternal: {fileID: 0}
-  serializedVersion: 5
+  m_CorrespondingSourceObject: {fileID: 0}
+  m_PrefabInstance: {fileID: 0}
+  m_PrefabAsset: {fileID: 0}
+  serializedVersion: 6
-  - component: {fileID: 1715640923}
   - component: {fileID: 1715640922}
   - component: {fileID: 1715640921}
   m_Layer: 0
 --- !u!81 &1715640921
 AudioListener:
   m_ObjectHideFlags: 0
-  m_PrefabParentObject: {fileID: 0}
-  m_PrefabInternal: {fileID: 0}
+  m_CorrespondingSourceObject: {fileID: 0}
+  m_PrefabInstance: {fileID: 0}
+  m_PrefabAsset: {fileID: 0}
-  m_PrefabParentObject: {fileID: 0}
-  m_PrefabInternal: {fileID: 0}
-  m_GameObject: {fileID: 1715640920}
-  m_Enabled: 1
---- !u!92 &1715640923
-Behaviour:
-  m_ObjectHideFlags: 0
-  m_PrefabParentObject: {fileID: 0}
-  m_PrefabInternal: {fileID: 0}
+  m_CorrespondingSourceObject: {fileID: 0}
+  m_PrefabInstance: {fileID: 0}
+  m_PrefabAsset: {fileID: 0}
-  m_PrefabParentObject: {fileID: 0}
-  m_PrefabInternal: {fileID: 0}
+  m_CorrespondingSourceObject: {fileID: 0}
+  m_PrefabInstance: {fileID: 0}
+  m_PrefabAsset: {fileID: 0}
+  m_projectionMatrixMode: 1
+  m_SensorSize: {x: 36, y: 24}
+  m_LensShift: {x: 0, y: 0}
+  m_GateFitMode: 2
+  m_FocalLength: 50
   m_NormalizedViewPortRect:
     serializedVersion: 2
     x: 0
 --- !u!4 &1715640925
 Transform:
   m_ObjectHideFlags: 0
-  m_PrefabParentObject: {fileID: 0}
-  m_PrefabInternal: {fileID: 0}
+  m_CorrespondingSourceObject: {fileID: 0}
+  m_PrefabInstance: {fileID: 0}
+  m_PrefabAsset: {fileID: 0}
   m_GameObject: {fileID: 1715640920}
   m_LocalRotation: {x: 0, y: 0, z: 0, w: 1}
   m_LocalPosition: {x: 0, y: 1, z: -10}

Project/Assets/ML-Agents/Examples/Template/Scripts/TemplateAgent.cs

     {
     }
 
-    public override void AgentAction(float[] vectorAction)
+    public override void OnActionReceived(float[] vectorAction)
-    public override void AgentReset()
+    public override void OnEpisodeBegin()
     {
     }
 }

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

 
     void Reset()
     {
-        m_AgentA.Done();
-        m_AgentB.Done();
+        m_AgentA.EndEpisode();
+        m_AgentB.EndEpisode();
         m_Area.MatchReset();
         lastFloorHit = FloorHit.Service;
         net = false;

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

 using UnityEngine.UI;
 using MLAgents;
 using MLAgents.Sensors;
+using MLAgents.SideChannels;
 
 public class TennisAgent : Agent
 {
     Rigidbody m_AgentRb;
     Rigidbody m_BallRb;
     float m_InvertMult;
-    IFloatProperties m_ResetParams;
+    FloatPropertiesChannel m_ResetParams;
 
     // Looks for the scoreboard based on the name of the gameObjects.
     // Do not modify the names of the Score GameObjects
 
-    public override void InitializeAgent()
+    public override void Initialize()
     {
         m_AgentRb = GetComponent<Rigidbody>();
         m_BallRb = ball.GetComponent<Rigidbody>();
         sensor.AddObservation(m_InvertMult * gameObject.transform.rotation.z);
     }
 
-    public override void AgentAction(float[] vectorAction)
+    public override void OnActionReceived(float[] vectorAction)
     {
         var moveX = Mathf.Clamp(vectorAction[0], -1f, 1f) * m_InvertMult;
         var moveY = Mathf.Clamp(vectorAction[1], -1f, 1f);
         return action;
     }
 
-    public override void AgentReset()
+    public override void OnEpisodeBegin()
     {
         m_InvertMult = invertX ? -1f : 1f;
 

Project/Assets/ML-Agents/Examples/Walker/Scripts/WalkerAgent.cs

 using MLAgents;
 using MLAgentsExamples;
 using MLAgents.Sensors;
+using MLAgents.SideChannels;
 
 public class WalkerAgent : Agent
 {
     Rigidbody m_ChestRb;
     Rigidbody m_SpineRb;
 
-    IFloatProperties m_ResetParams;
+    FloatPropertiesChannel m_ResetParams;
-    public override void InitializeAgent()
+    public override void Initialize()
     {
         m_JdController = GetComponent<JointDriveController>();
         m_JdController.SetupBodyPart(hips);
         }
     }
 
-    public override void AgentAction(float[] vectorAction)
+    public override void OnActionReceived(float[] vectorAction)
     {
         var bpDict = m_JdController.bodyPartsDict;
         var i = -1;
     /// <summary>
     /// Loop over body parts and reset them to initial conditions.
     /// </summary>
-    public override void AgentReset()
+    public override void OnEpisodeBegin()
     {
         if (m_DirToTarget != Vector3.zero)
         {

Project/Assets/ML-Agents/Examples/WallJump/Scripts/WallJumpAgent.cs

     Vector3 m_JumpTargetPos;
     Vector3 m_JumpStartingPos;
 
-    public override void InitializeAgent()
+    public override void Initialize()
     {
         m_WallJumpSettings = FindObjectOfType<WallJumpSettings>();
         m_Configuration = Random.Range(0, 5);
         jumpingTime -= Time.fixedDeltaTime;
     }
 
-    public override void AgentAction(float[] vectorAction)
+    public override void OnActionReceived(float[] vectorAction)
     {
         MoveAgent(vectorAction);
         if ((!Physics.Raycast(m_AgentRb.position, Vector3.down, 20))
-            Done();
+            EndEpisode();
             ResetBlock(m_ShortBlockRb);
             StartCoroutine(
                 GoalScoredSwapGroundMaterial(m_WallJumpSettings.failMaterial, .5f));
         if (col.gameObject.CompareTag("goal") && DoGroundCheck(true))
         {
             SetReward(1f);
-            Done();
+            EndEpisode();
             StartCoroutine(
                 GoalScoredSwapGroundMaterial(m_WallJumpSettings.goalScoredMaterial, 2));
         }
         blockRb.angularVelocity = Vector3.zero;
     }
 
-    public override void AgentReset()
+    public override void OnEpisodeBegin()
     {
         ResetBlock(m_ShortBlockRb);
         transform.localPosition = new Vector3(
                 Academy.Instance.FloatProperties.GetPropertyWithDefault("no_wall_height", 0),
                 localScale.z);
             wall.transform.localScale = localScale;
-            GiveModel("SmallWallJump", noWallBrain);
+            SetModel("SmallWallJump", noWallBrain);
         }
         else if (config == 1)
         {
                 localScale.z);
             wall.transform.localScale = localScale;
-            GiveModel("SmallWallJump", smallWallBrain);
+            SetModel("SmallWallJump", smallWallBrain);
         }
         else
         {
                 height,
                 localScale.z);
             wall.transform.localScale = localScale;
-            GiveModel("BigWallJump", bigWallBrain);
+            SetModel("BigWallJump", bigWallBrain);
         }
     }
 }

Project/Packages/manifest.json

 {
   "dependencies": {
     "com.unity.ads": "2.0.8",
-    "com.unity.analytics": "3.2.2",
+    "com.unity.analytics": "3.2.3",
     "com.unity.collab-proxy": "1.2.15",
     "com.unity.ml-agents": "file:../../com.unity.ml-agents",
     "com.unity.package-manager-ui": "2.0.8",
     "com.unity.modules.wind": "1.0.0",
     "com.unity.modules.xr": "1.0.0"
   },
-  "testables" : [
+  "testables": [
     "com.unity.ml-agents"
   ]
 }

Project/ProjectSettings/ProjectSettings.asset

   useOnDemandResources: 0
   accelerometerFrequency: 60
   companyName: Unity Technologies
-  productName: Unity Environment
+  productName: UnityEnvironment
-  m_ShowUnitySplashScreen: 0
+  m_ShowUnitySplashScreen: 1
   m_ShowUnitySplashLogo: 1
   m_SplashScreenOverlayOpacity: 1
   m_SplashScreenAnimation: 1
   xboxOneMonoLoggingLevel: 0
   xboxOneLoggingLevel: 1
   xboxOneDisableEsram: 0
+  xboxOneEnableTypeOptimization: 0
   xboxOnePresentImmediateThreshold: 0
   switchQueueCommandMemory: 1048576
   switchQueueControlMemory: 16384

Project/ProjectSettings/ProjectVersion.txt

-m_EditorVersion: 2018.4.14f1
+m_EditorVersion: 2018.4.18f1

README.md

 used for multiple purposes, including controlling NPC behavior (in a variety of
 settings such as multi-agent and adversarial), automated testing of game builds
 and evaluating different game design decisions pre-release. The ML-Agents
-toolkit is mutually beneficial for both game developers and AI researchers as it
+Toolkit is mutually beneficial for both game developers and AI researchers as it
 provides a central platform where advances in AI can be evaluated on Unity’s
 rich environments and then made accessible to the wider research and game
 developer communities.
 * Unity environment control from Python
-* 10+ sample Unity environments
+* 15+ sample Unity environments
 * Two deep reinforcement learning algorithms,
 [Proximal Policy Optimization](https://github.com/Unity-Technologies/ml-agents/tree/latest_release/docs/Training-PPO.md)
  (PPO) and [Soft Actor-Critic](https://github.com/Unity-Technologies/ml-agents/tree/latest_release/docs/Training-SAC.md)
 * Train memory-enhanced agents using deep reinforcement learning
 * Easily definable Curriculum Learning and Generalization scenarios
-* Built-in support for Imitation Learning
+* Built-in support for [Imitation Learning](https://github.com/Unity-Technologies/ml-agents/tree/latest_release/docs/Training-Imitation-Learning.md) through Behavioral Cloning or Generative Adversarial Imitation Learning
-* Simplified set-up with Docker
-## Documentation
+## Releases & Documentation
+**Our latest, stable release is 0.14.1. Click
+[here](https://github.com/Unity-Technologies/ml-agents/tree/latest_release/docs/Readme.md) to
+get started with the latest release of ML-Agents.**
+
+The table below lists all our releases, including our `master` branch which is under active
+development and may be unstable. A few helpful guidelines:
+* The docs links in the table below include installation and usage instructions specific to each
+release. Remember to always use the documentation that corresponds to the release version you're
+using.
+* See the [GitHub releases](https://github.com/Unity-Technologies/ml-agents/releases) for more
+details of the changes between versions.
+* If you have used an earlier version of the ML-Agents Toolkit, we strongly recommend our
+[guide on migrating from earlier versions](docs/Migrating.md).
+
+
+| **Version** | **Release Date** | **Source** | **Documentation** | **Download** |
+|:-------:|:------:|:-------------:|:-------:|:------------:|
+| **master** (unstable) | -- | [source](https://github.com/Unity-Technologies/ml-agents/tree/master) |  [docs](https://github.com/Unity-Technologies/ml-agents/tree/master/docs/Readme.md) | [download](https://github.com/Unity-Technologies/ml-agents/archive/master.zip) |
+| **0.14.1** (latest stable release) | February 26, 2020 | **[source](https://github.com/Unity-Technologies/ml-agents/tree/latest_release)** |  **[docs](https://github.com/Unity-Technologies/ml-agents/tree/latest_release/docs/Readme.md)** | **[download](https://github.com/Unity-Technologies/ml-agents/archive/latest_release.zip)** |
+| **0.14.0**  | February 13, 2020 | [source](https://github.com/Unity-Technologies/ml-agents/tree/0.14.0) |  [docs](https://github.com/Unity-Technologies/ml-agents/tree/0.14.0/docs/Readme.md) | [download](https://github.com/Unity-Technologies/ml-agents/archive/0.14.0.zip) |
+| **0.13.1**  | January 21, 2020 | [source](https://github.com/Unity-Technologies/ml-agents/tree/0.13.1) |  [docs](https://github.com/Unity-Technologies/ml-agents/tree/0.13.1/docs/Readme.md) | [download](https://github.com/Unity-Technologies/ml-agents/archive/0.13.1.zip) |
+| **0.13.0**  | January 8, 2020 | [source](https://github.com/Unity-Technologies/ml-agents/tree/0.13.0) |  [docs](https://github.com/Unity-Technologies/ml-agents/tree/0.13.0/docs/Readme.md) | [download](https://github.com/Unity-Technologies/ml-agents/archive/0.13.0.zip) |
+| **0.12.1**  | December 11, 2019 | [source](https://github.com/Unity-Technologies/ml-agents/tree/0.12.1) |  [docs](https://github.com/Unity-Technologies/ml-agents/tree/0.12.1/docs/Readme.md) | [download](https://github.com/Unity-Technologies/ml-agents/archive/0.12.1.zip) |
+| **0.12.0**  | December 2, 2019 | [source](https://github.com/Unity-Technologies/ml-agents/tree/0.12.0) |  [docs](https://github.com/Unity-Technologies/ml-agents/tree/0.12.0/docs/Readme.md) | [download](https://github.com/Unity-Technologies/ml-agents/archive/0.12.0.zip) |
+| **0.11.0**  | November 4, 2019 | [source](https://github.com/Unity-Technologies/ml-agents/tree/0.11.0) |  [docs](https://github.com/Unity-Technologies/ml-agents/tree/0.11.0/docs/Readme.md) | [download](https://github.com/Unity-Technologies/ml-agents/archive/0.11.0.zip) |
+| **0.10.1**  | October 9, 2019 | [source](https://github.com/Unity-Technologies/ml-agents/tree/0.10.1) |  [docs](https://github.com/Unity-Technologies/ml-agents/tree/0.10.1/docs/Readme.md) | [download](https://github.com/Unity-Technologies/ml-agents/archive/0.10.1.zip) |
+| **0.10.0**  | September 30, 2019 | [source](https://github.com/Unity-Technologies/ml-agents/tree/0.10.0) |  [docs](https://github.com/Unity-Technologies/ml-agents/tree/0.10.0/docs/Readme.md) | [download](https://github.com/Unity-Technologies/ml-agents/archive/0.10.0.zip) |
+
+
+## Citation
+
+If you are a researcher interested in a discussion of Unity as an AI platform, see a pre-print
+of our [reference paper on Unity and the ML-Agents Toolkit](https://arxiv.org/abs/1809.02627).
+
+If you use Unity or the ML-Agents Toolkit to conduct research, we ask that you cite the following
+paper as a reference:
+
+Juliani, A., Berges, V., Vckay, E., Gao, Y., Henry, H., Mattar, M., Lange, D. (2018). Unity: A General Platform for Intelligent Agents. *arXiv preprint arXiv:1809.02627.* https://github.com/Unity-Technologies/ml-agents.
-* For more information, in addition to installation and usage instructions, see
-  the [documentation for the latest release](https://github.com/Unity-Technologies/ml-agents/tree/latest_release/docs/Readme.md).
-* If you are a researcher interested in a discussion of Unity as an AI platform, see a pre-print of our [reference paper on Unity and the ML-Agents Toolkit](https://arxiv.org/abs/1809.02627). Also, see below for instructions on citing this paper.
-* If you have used an earlier version of the ML-Agents toolkit, we strongly
-  recommend our [guide on migrating from earlier versions](docs/Migrating.md).
-
+* (February 28, 2020) [Training intelligent adversaries using self-play with ML-Agents](https://blogs.unity3d.com/2020/02/28/training-intelligent-adversaries-using-self-play-with-ml-agents/)
+* (November 11, 2019) [Training your agents 7 times faster with ML-Agents](https://blogs.unity3d.com/2019/11/11/training-your-agents-7-times-faster-with-ml-agents/)
+* (October 21, 2019) [The AI@Unity interns help shape the world](https://blogs.unity3d.com/2019/10/21/the-aiunity-interns-help-shape-the-world/)
+* (April 15, 2019) [Unity ML-Agents Toolkit v0.8: Faster training on real games](https://blogs.unity3d.com/2019/04/15/unity-ml-agents-toolkit-v0-8-faster-training-on-real-games/)
+* (March 1, 2019) [Unity ML-Agents Toolkit v0.7: A leap towards cross-platform inference](https://blogs.unity3d.com/2019/03/01/unity-ml-agents-toolkit-v0-7-a-leap-towards-cross-platform-inference/)
+* (December 17, 2018) [ML-Agents Toolkit v0.6: Improved usability of Brains and Imitation Learning](https://blogs.unity3d.com/2018/12/17/ml-agents-toolkit-v0-6-improved-usability-of-brains-and-imitation-learning/)
+* (October 2, 2018) [Puppo, The Corgi: Cuteness Overload with the Unity ML-Agents Toolkit](https://blogs.unity3d.com/2018/10/02/puppo-the-corgi-cuteness-overload-with-the-unity-ml-agents-toolkit/)
+* (September 11, 2018) [ML-Agents Toolkit v0.5, new resources for AI researchers available now](https://blogs.unity3d.com/2018/09/11/ml-agents-toolkit-v0-5-new-resources-for-ai-researchers-available-now/)
+* (June 26, 2018) [Solving sparse-reward tasks with Curiosity](https://blogs.unity3d.com/2018/06/26/solving-sparse-reward-tasks-with-curiosity/)
+* (June 19, 2018) [Unity ML-Agents Toolkit v0.4 and Udacity Deep Reinforcement Learning Nanodegree](https://blogs.unity3d.com/2018/06/19/unity-ml-agents-toolkit-v0-4-and-udacity-deep-reinforcement-learning-nanodegree/)
+* (May 24, 2018) [Imitation Learning in Unity: The Workflow](https://blogs.unity3d.com/2018/05/24/imitation-learning-in-unity-the-workflow/)
+* (March 15, 2018) [ML-Agents Toolkit v0.3 Beta released: Imitation Learning, feedback-driven features, and more](https://blogs.unity3d.com/2018/03/15/ml-agents-v0-3-beta-released-imitation-learning-feedback-driven-features-and-more/)
+* (December 11, 2017) [Using Machine Learning Agents in a real game: a beginner’s guide](https://blogs.unity3d.com/2017/12/11/using-machine-learning-agents-in-a-real-game-a-beginners-guide/)
+* (December 8, 2017) [Introducing ML-Agents Toolkit v0.2: Curriculum Learning, new environments, and more](https://blogs.unity3d.com/2017/12/08/introducing-ml-agents-v0-2-curriculum-learning-new-environments-and-more/)
+* (September 19, 2017) [Introducing: Unity Machine Learning Agents Toolkit](https://blogs.unity3d.com/2017/09/19/introducing-unity-machine-learning-agents/)
-* [Using Machine Learning Agents in a real game: a beginner’s guide](https://blogs.unity3d.com/2017/12/11/using-machine-learning-agents-in-a-real-game-a-beginners-guide/)
-* [Post](https://blogs.unity3d.com/2018/02/28/introducing-the-winners-of-the-first-ml-agents-challenge/)
-  announcing the winners of our
-  [first ML-Agents Challenge](https://connect.unity.com/challenges/ml-agents-1)
-* [Post](https://blogs.unity3d.com/2018/01/23/designing-safer-cities-through-simulations/)
-  overviewing how Unity can be leveraged as a simulator to design safer cities.
 
 In addition to our own documentation, here are some additional, relevant articles:
 
 
 ## Community and Feedback
 
-The ML-Agents toolkit is an open-source project and we encourage and welcome
+The ML-Agents Toolkit is an open-source project and we encourage and welcome
-For problems with the installation and setup of the the ML-Agents toolkit, or
+For problems with the installation and setup of the the ML-Agents Toolkit, or
-If you run into any other problems using the ML-Agents toolkit, or have a specific
+If you run into any other problems using the ML-Agents Toolkit, or have a specific
-Your opinion matters a great deal to us. Only by hearing your thoughts on the Unity ML-Agents Toolkit can we continue
-to improve and grow. Please take a few minutes to [let us know about it](https://github.com/Unity-Technologies/ml-agents/issues/1454).
+Your opinion matters a great deal to us. Only by hearing your thoughts on the Unity ML-Agents
+Toolkit can we continue to improve and grow. Please take a few minutes to
+[let us know about it](https://github.com/Unity-Technologies/ml-agents/issues/1454).
-## Releases
-The latest release is 0.14.1. Previous releases can be found below:
-
-| **Version** | **Source** | **Documentation** | **Download** |
-|:-------:|:------:|:-------------:|:-------:|
-| **0.14.0**  | [source](https://github.com/Unity-Technologies/ml-agents/tree/0.14.0) |  [docs](https://github.com/Unity-Technologies/ml-agents/tree/0.14.0/docs) | [download](https://github.com/Unity-Technologies/ml-agents/archive/0.14.0.zip) |
-| **0.13.1**  | [source](https://github.com/Unity-Technologies/ml-agents/tree/0.13.1) |  [docs](https://github.com/Unity-Technologies/ml-agents/tree/0.13.1/docs) | [download](https://github.com/Unity-Technologies/ml-agents/archive/0.13.1.zip) |
-| **0.13.0**  | [source](https://github.com/Unity-Technologies/ml-agents/tree/0.13.0) |  [docs](https://github.com/Unity-Technologies/ml-agents/tree/0.13.0/docs) | [download](https://github.com/Unity-Technologies/ml-agents/archive/0.13.0.zip) |
-| **0.12.1**  | [source](https://github.com/Unity-Technologies/ml-agents/tree/0.12.1) |  [docs](https://github.com/Unity-Technologies/ml-agents/tree/0.12.1/docs) | [download](https://github.com/Unity-Technologies/ml-agents/archive/0.12.1.zip) |
-| **0.12.0**  | [source](https://github.com/Unity-Technologies/ml-agents/tree/0.12.0) |  [docs](https://github.com/Unity-Technologies/ml-agents/tree/0.12.0/docs) | [download](https://github.com/Unity-Technologies/ml-agents/archive/0.12.0.zip) |
-| **0.11.0**  | [source](https://github.com/Unity-Technologies/ml-agents/tree/0.11.0) |  [docs](https://github.com/Unity-Technologies/ml-agents/tree/0.11.0/docs) | [download](https://github.com/Unity-Technologies/ml-agents/archive/0.11.0.zip) |
-| **0.10.1**  | [source](https://github.com/Unity-Technologies/ml-agents/tree/0.10.1) |  [docs](https://github.com/Unity-Technologies/ml-agents/tree/0.10.1/docs) | [download](https://github.com/Unity-Technologies/ml-agents/archive/0.10.1.zip) |
-| **0.10.0**  | [source](https://github.com/Unity-Technologies/ml-agents/tree/0.10.0) |  [docs](https://github.com/Unity-Technologies/ml-agents/tree/0.10.0/docs) | [download](https://github.com/Unity-Technologies/ml-agents/archive/0.10.0.zip) |
-
-See the [GitHub releases](https://github.com/Unity-Technologies/ml-agents/releases) for more details of the changes
-between versions.
-
-Please note that the `master` branch is under active development, so the documentation there may differ from the code
-of a previous release. Always use the documentation that corresponds to the release version you're using.
-
-## Citation
-
-If you use Unity or the ML-Agents Toolkit to conduct research, we ask that you cite the following paper as a reference:
-
-Juliani, A., Berges, V., Vckay, E., Gao, Y., Henry, H., Mattar, M., Lange, D. (2018). Unity: A General Platform for Intelligent Agents. *arXiv preprint arXiv:1809.02627.* https://github.com/Unity-Technologies/ml-agents.

com.unity.ml-agents/CHANGELOG.md

 
 ## [Unreleased]
 ### Major Changes
+
+### Minor Changes
+ - Format of console output has changed slightly and now matches the name of the model/summary directory. (#3630, #3616)
+
+## [0.15.0-preview] - 2020-03-18
+### Major Changes
  - `Agent.CollectObservations` now takes a VectorSensor argument. (#3352, #3389)
  - Added `Agent.CollectDiscreteActionMasks` virtual method with a `DiscreteActionMasker` argument to specify which discrete actions are unavailable to the Agent. (#3525)
  - Beta support for ONNX export was added. If the `tf2onnx` python package is installed, models will be saved to `.onnx` as well as `.nn` format.
+ - All SideChannel related code has been moved to the namespace `MLAgents.SideChannels`.
+ - `BrainParameters` and `SpaceType` have been removed from the public API
+ - `BehaviorParameters` have been removed from the public API.
+ - The following methods in the `Agent` class have been deprecated and will be removed in a later release:
+   - `InitializeAgent()` was renamed to `Initialize()`
+   - `AgentAction()` was renamed to `OnActionReceived()`
+   - `AgentReset()` was renamed to `OnEpisodeBegin()`
+   - `Done()` was renamed to `EndEpisode()`
+   - `GiveModel()` was renamed to `SetModel()`
 
 ### Minor Changes
  - Monitor.cs was moved to Examples. (#3372)
  - Academy.RegisterSideChannel and UnregisterSideChannel methods were added. (#3391)
  - A tutorial on adding custom SideChannels was added (#3391)
  - The stepping logic for the Agent and the Academy has been simplified (#3448)
- - Update Barracuda to 0.6.0-preview
+ - Update Barracuda to 0.6.1-preview
  * The interface for `RayPerceptionSensor.PerceiveStatic()` was changed to take an input class and write to an output class, and the method was renamed to `Perceive()`.
  - The checkpoint file suffix was changed from `.cptk` to `.ckpt` (#3470)
  - The command-line argument used to determine the port that an environment will listen on was changed from `--port` to `--mlagents-port`.
  - The method `GetStepCount()` on the Agent class has been replaced with the property getter `StepCount`
  - `RayPerceptionSensorComponent` and related classes now display the debug gizmos whenever the Agent is selected (not just Play mode).
  - Most fields on `RayPerceptionSensorComponent` can now be changed while the editor is in Play mode. The exceptions to this are fields that affect the number of observations.
+ - Most fields on `CameraSensorComponent` and `RenderTextureSensorComponent` were changed to private and replaced by properties with the same name.
-
+ - The `Agent` class is no longer abstract.
+ - SensorBase was moved out of the package and into the Examples directory.
+ - `AgentInfo.actionMasks` has been renamed to `AgentInfo.discreteActionMasks`.
+ - `DecisionRequester` has been made internal (you can still use the DecisionRequesterComponent from the inspector). `RepeatAction` was renamed `TakeActionsBetweenDecisions` for clarity. (#3555)
+ - The `IFloatProperties` interface has been removed.
+ - Fix #3579.
+ - Improved inference performance for models with multiple action branches. (#3598)
+ - Fixed an issue when using GAIL with less than `batch_size` number of demonstrations. (#3591)
+ - The interfaces to the `SideChannel` classes (on C# and python) have changed to use new  `IncomingMessage` and `OutgoingMessage` classes. These should make reading and writing data to the channel easier. (#3596)
+ - Updated the ExpertPyramid.demo example demonstration file (#3613)
+ - Updated project version for example environments to 2018.4.18f1. (#3618)
+ - Changed the Product Name in the example environments to remove spaces, so that the default build executable file doesn't contain spaces. (#3612)
 
 ## [0.14.1-preview] - 2020-02-25
 

com.unity.ml-agents/Documentation~/com.unity.ml-agents.md

-Please see the [ML-Agents README)(https://github.com/Unity-Technologies/ml-agents/blob/master/README.md)
+# About ML-Agents package (`com.unity.ml-agents`)
+
+The Unity ML-Agents package contains the C# SDK for the
+[Unity ML-Agents Toolkit](https://github.com/Unity-Technologies/ml-agents).
+
+The package provides the ability for any Unity scene to be converted into a learning
+environment where character behaviors can be trained using a variety of machine learning
+algorithms. Additionally, it enables any trained behavior to be embedded back into the Unity
+scene. More specifically, the package provides the following core functionalities:
+* Define Agents: entities whose behavior will be learned. Agents are entities
+that generate observations (through sensors), take actions and receive rewards from
+the environment.
+* Define Behaviors: entities that specifiy how an agent should act. Multiple agents can
+share the same Behavior and a scene may have multiple Behaviors.
+* Record demonstrations of an agent within the Editor. These demonstrations can be
+valuable to train a behavior for that agent.
+* Embedding a trained behavior into the scene via the
+[Unity Inference Engine](https://docs.unity3d.com/Packages/com.unity.barracuda@latest/index.html).
+Thus an Agent can switch from a learning behavior to an inference behavior.
+
+Note that this package does not contain the machine learning algorithms for training
+behaviors. It relies on a Python package to orchestrate the training. This package
+only enables instrumenting a Unity scene and setting it up for training, and then
+embedding the trained model back into your Unity scene.
+
+## Preview package
+This package is available as a preview, so it is not ready for production use.
+The features and documentation in this package might change before it is verified for release.
+
+
+## Package contents
+
+The following table describes the package folder structure:
+
+|**Location**|**Description**|
+|---|---|
+|*Documentation~*|Contains the documentation for the Unity package.|
+|*Editor*|Contains utilities for Editor windows and drawers.|
+|*Plugins*|Contains third-party DLLs.|
+|*Runtime*|Contains core C# APIs for integrating ML-Agents into your Unity scene. |
+|*Tests*|Contains the unit tests for the package.|
+
+<a name="Installation"></a>
+
+## Installation
+
+To install this package, follow the instructions in the
+[Package Manager documentation](https://docs.unity3d.com/Manual/upm-ui-install.html).
+
+To install the Python package to enable training behaviors, follow the instructions on our
+[GitHub repository](https://github.com/Unity-Technologies/ml-agents/blob/latest_release/docs/Installation.md).
+
+## Requirements
+
+This version of the Unity ML-Agents package is compatible with the following versions of the Unity Editor:
+
+* 2018.4 and later (recommended)
+
+## Known limitations
+
+### Headless Mode
+
+If you enable Headless mode, you will not be able to collect visual observations
+from your agents.
+
+### Rendering Speed and Synchronization
+
+Currently the speed of the game physics can only be increased to 100x real-time.
+The Academy also moves in time with FixedUpdate() rather than Update(), so game
+behavior implemented in Update() may be out of sync with the agent decision
+making. See
+[Execution Order of Event Functions](https://docs.unity3d.com/Manual/ExecutionOrder.html)
+for more information.
+
+You can control the frequency of Academy stepping by calling
+`Academy.Instance.DisableAutomaticStepping()`, and then calling
+`Academy.Instance.EnvironmentStep()`
+
+### Unity Inference Engine Models
+Currently, only models created with our trainers are supported for running
+ML-Agents with a neural network behavior.
+
+
+## Helpful links
+
+If you are new to the Unity ML-Agents package, or have a question after reading
+the documentation, you can checkout our
+[GitHUb Repository](https://github.com/Unity-Technologies/ml-agents), which
+also includes a number of ways to
+[connect with us](https://github.com/Unity-Technologies/ml-agents#community-and-feedback)
+including our [ML-Agents Forum](https://forum.unity.com/forums/ml-agents.453/).
+

com.unity.ml-agents/Editor/BehaviorParametersEditor.cs

 using MLAgents.Sensors;
 using UnityEditor;
 using Barracuda;
+using MLAgents.Policies;
+using UnityEngine;
 
 namespace MLAgents.Editor
 {
         {
             var so = serializedObject;
             so.Update();
+            bool needPolicyUpdate; // Whether the name, model, inference device, or BehaviorType changed.
+            EditorGUI.indentLevel++;
+            EditorGUI.BeginChangeCheck(); // global
+
+            EditorGUI.BeginChangeCheck();
+            {
+                EditorGUILayout.PropertyField(so.FindProperty("m_BehaviorName"));
+            }
+            needPolicyUpdate = EditorGUI.EndChangeCheck();
+
+            EditorGUI.BeginDisabledGroup(!EditorUtilities.CanUpdateModelProperties());
+            {
+                EditorGUILayout.PropertyField(so.FindProperty("m_BrainParameters"), true);
+            }
+            EditorGUI.EndDisabledGroup();
+
-            EditorGUI.indentLevel++;
-            EditorGUILayout.PropertyField(so.FindProperty("m_BehaviorName"));
-            EditorGUILayout.PropertyField(so.FindProperty("m_BrainParameters"), true);
-            EditorGUILayout.PropertyField(so.FindProperty("m_Model"), true);
-            EditorGUI.indentLevel++;
-            EditorGUILayout.PropertyField(so.FindProperty("m_InferenceDevice"), true);
-            EditorGUI.indentLevel--;
-            EditorGUILayout.PropertyField(so.FindProperty("m_BehaviorType"));
-            EditorGUILayout.PropertyField(so.FindProperty("m_TeamID"));
-            EditorGUILayout.PropertyField(so.FindProperty("m_UseChildSensors"), true);
-            // EditorGUILayout.PropertyField(serializedObject.FindProperty("m_Heuristic"), true);
-            EditorGUI.indentLevel--;
-            if (EditorGUI.EndChangeCheck())
+            {
+                EditorGUILayout.PropertyField(so.FindProperty("m_Model"), true);
+                EditorGUI.indentLevel++;
+                EditorGUILayout.PropertyField(so.FindProperty("m_InferenceDevice"), true);
+                EditorGUI.indentLevel--;
+            }
+            needPolicyUpdate = needPolicyUpdate || EditorGUI.EndChangeCheck();
+
+            EditorGUI.BeginChangeCheck();
+            {
+                EditorGUILayout.PropertyField(so.FindProperty("m_BehaviorType"));
+            }
+            needPolicyUpdate = needPolicyUpdate || EditorGUI.EndChangeCheck();
+
+            EditorGUILayout.PropertyField(so.FindProperty("TeamId"));
+            EditorGUI.BeginDisabledGroup(!EditorUtilities.CanUpdateModelProperties());
-                m_RequireReload = true;
+                EditorGUILayout.PropertyField(so.FindProperty("m_UseChildSensors"), true);
+            EditorGUI.EndDisabledGroup();
+
+            EditorGUI.indentLevel--;
+            m_RequireReload = EditorGUI.EndChangeCheck();
+
+            if (needPolicyUpdate)
+            {
+                UpdateAgentPolicy();
+            }
         }
 
         /// <summary>
             }
             if (brainParameters != null)
             {
-                var failedChecks = InferenceBrain.BarracudaModelParamLoader.CheckModel(
-                    barracudaModel, brainParameters, sensorComponents);
+                var failedChecks = Inference.BarracudaModelParamLoader.CheckModel(
+                    barracudaModel, brainParameters, sensorComponents, behaviorParameters.behaviorType
+                );
                 foreach (var check in failedChecks)
                 {
                     if (check != null)
                 }
             }
+        }
+
+        void UpdateAgentPolicy()
+        {
+            var behaviorParameters = (BehaviorParameters)target;
+            behaviorParameters.UpdateAgentPolicy();
         }
     }
 }

com.unity.ml-agents/Editor/BrainParametersDrawer.cs

 using UnityEngine;
 using UnityEditor;
+using MLAgents.Policies;
 
 namespace MLAgents.Editor
 {

com.unity.ml-agents/Editor/DemonstrationDrawer.cs

 using System.Text;
-using MLAgents;
+using MLAgents.Policies;
 
 
 namespace MLAgents.Editor

com.unity.ml-agents/Editor/RayPerceptionSensorComponentBaseEditor.cs

             EditorGUI.BeginChangeCheck();
             EditorGUI.indentLevel++;
 
-            EditorGUILayout.PropertyField(so.FindProperty("m_SensorName"), true);
-
-            // Because the number of rays and the tags affect the observation shape,
-            // they are not editable during play mode.
-            EditorGUI.BeginDisabledGroup(Application.isPlaying);
+            // Don't allow certain fields to be modified during play mode.
+            // * SensorName affects the ordering of the Agent's observations
+            // * The number of tags and rays affects the size of the observations.
+            EditorGUI.BeginDisabledGroup(!EditorUtilities.CanUpdateModelProperties());
+                EditorGUILayout.PropertyField(so.FindProperty("m_SensorName"), true);
                 EditorGUILayout.PropertyField(so.FindProperty("m_DetectableTags"), true);
                 EditorGUILayout.PropertyField(so.FindProperty("m_RaysPerDirection"), true);
             }
 
             // Because the number of observation stacks affects the observation shape,
             // it is not editable during play mode.
-            EditorGUI.BeginDisabledGroup(Application.isPlaying);
+            EditorGUI.BeginDisabledGroup(!EditorUtilities.CanUpdateModelProperties());
             {
                 EditorGUILayout.PropertyField(so.FindProperty("m_ObservationStacks"), true);
             }
                 m_RequireSensorUpdate = true;
             }
 
-            UpdateSensorIfDirty();
+            UpdateSensorIfDirty();
-
-                if (m_RequireSensorUpdate)
-                {
-                    var sensorComponent = serializedObject.targetObject as RayPerceptionSensorComponentBase;
-                    sensorComponent?.UpdateSensor();
-                    m_RequireSensorUpdate = false;
-                }
+            if (m_RequireSensorUpdate)
+            {
+                var sensorComponent = serializedObject.targetObject as RayPerceptionSensorComponentBase;
+                sensorComponent?.UpdateSensor();
+                m_RequireSensorUpdate = false;
+            }
         }
     }
 

com.unity.ml-agents/Runtime/Academy.cs

 #if UNITY_EDITOR
 using UnityEditor;
 #endif
-using MLAgents.InferenceBrain;
+using MLAgents.Inference;
+using MLAgents.Policies;
+using MLAgents.SideChannels;
 using Barracuda;
 
 /**
         "docs/Learning-Environment-Design.md")]
     public class Academy : IDisposable
     {
-        const string k_ApiVersion = "API-15-dev0";
+        /// <summary>
+        /// Communication protocol version.
+        /// When connecting to python, this must match UnityEnvironment.API_VERSION.
+        /// Currently we require strict equality between the communication protocol
+        /// on each side, although we may allow some flexibility in the future.
+        /// This should be incremented whenever a change is made to the communication protocol.
+        /// </summary>
+        const string k_ApiVersion = "0.15.0";
+
+        /// <summary>
+        /// Unity package version of com.unity.ml-agents.
+        /// This must match the version string in package.json and is checked in a unit test.
+        /// </summary>
+        internal const string k_PackageVersion = "0.15.0-preview";
+
-        internal const string k_portCommandLineFlag = "--mlagents-port";
+
+        const string k_portCommandLineFlag = "--mlagents-port";
 
         // Lazy initializer pattern, see https://csharpindepth.com/articles/singleton#lazy
         static Lazy<Academy> s_Lazy = new Lazy<Academy>(() => new Academy());
         /// <summary>
         /// Collection of float properties (indexed by a string).
         /// </summary>
-        public IFloatProperties FloatProperties;
+        public FloatPropertiesChannel FloatProperties;
 
 
         // Fields not provided in the Inspector.
         List<ModelRunner> m_ModelRunners = new List<ModelRunner>();
 
         // Flag used to keep track of the first time the Academy is reset.
-        bool m_FirstAcademyReset;
+        bool m_HadFirstReset;
 
         // The Academy uses a series of events to communicate with agents
         // to facilitate synchronization. More specifically, it ensure
                     var unityRlInitParameters = Communicator.Initialize(
                         new CommunicatorInitParameters
                         {
-                            version = k_ApiVersion,
+                            unityCommunicationVersion = k_ApiVersion,
+                            unityPackageVersion = k_PackageVersion,
                             name = "AcademySingleton",
                         });
                     UnityEngine.Random.InitState(unityRlInitParameters.seed);
         {
             EnvironmentReset();
             AgentForceReset?.Invoke();
-            m_FirstAcademyReset = true;
+            m_HadFirstReset = true;
         }
 
         /// <summary>
         public void EnvironmentStep()
         {
-            if (!m_FirstAcademyReset)
+            if (!m_HadFirstReset)
             {
                 ForcedFullReset();
             }
         public void Dispose()
         {
             DisableAutomaticStepping();
+
             // Signal to listeners that the academy is being destroyed now
             DestroyAction?.Invoke();
 
                 {
                     mr.Dispose();
                 }
+
                 m_ModelRunners = null;
             }
 

com.unity.ml-agents/Runtime/Agent.cs

 using Barracuda;
 using MLAgents.Sensors;
 using MLAgents.Demonstrations;
+using MLAgents.Policies;
 
 namespace MLAgents
 {
         /// For discrete control, specifies the actions that the agent cannot take. Is true if
         /// the action is masked.
         /// </summary>
-        public bool[] actionMasks;
+        public bool[] discreteActionMasks;
 
         /// <summary>
         /// Current agent reward.
     /// value takes precedence (since the agent max step will never be reached).
     ///
     /// Lastly, note that at any step the policy to the agent is allowed to
-    /// change model with <see cref="GiveModel"/>.
+    /// change model with <see cref="SetModel"/>.
     ///
     /// Implementation-wise, it is required that this class is extended and the
     /// virtual methods overridden. For sample implementations of agent behavior,
         "docs/Learning-Environment-Design-Agents.md")]
     [Serializable]
     [RequireComponent(typeof(BehaviorParameters))]
-    public abstract class Agent : MonoBehaviour, ISerializationCallbackReceiver
+    public class Agent : MonoBehaviour, ISerializationCallbackReceiver
     {
         IPolicy m_Brain;
         BehaviorParameters m_PolicyFactory;
         /// </summary>
         internal VectorSensor collectObservationsSensor;
 
-        void OnEnable()
+        /// <summary>
+        /// Called when the attached <see cref="GameObject"/> becomes enabled and active.
+        /// </summary>
+        protected virtual void OnEnable()
         {
             LazyInitialize();
         }
             Academy.Instance.AgentForceReset += _AgentReset;
             m_Brain = m_PolicyFactory.GeneratePolicy(Heuristic);
             ResetData();
-            InitializeAgent();
+            Initialize();
+
+            // The first time the Academy resets, all Agents in the scene will be
+            // forced to reset through the <see cref="AgentForceReset"/> event.
+            // To avoid the Agent resetting twice, the Agents will not begin their
+            // episode when initializing until after the Academy had its first reset.
+            if (Academy.Instance.TotalStepCount != 0)
+            {
+                OnEpisodeBegin();
+            }
         }
 
         /// <summary>
             Disabled,
         }
 
-        void OnDisable()
+        /// <summary>
+        /// Called when the attached <see cref="GameObject"/> becomes disabled and inactive.
+        /// </summary>
+        protected virtual void OnDisable()
         {
             DemonstrationWriters.Clear();
 
             m_Brain?.RequestDecision(m_Info, sensors);
 
             // We also have to write any to any DemonstationStores so that they get the "done" flag.
-            foreach(var demoWriter in DemonstrationWriters)
+            foreach (var demoWriter in DemonstrationWriters)
             {
                 demoWriter.Record(m_Info, sensors);
             }
             m_RequestDecision = false;
         }
 
+        [Obsolete("GiveModel() has been deprecated, use SetModel() instead.")]
+        public void GiveModel(
+            string behaviorName,
+            NNModel model,
+            InferenceDevice inferenceDevice = InferenceDevice.CPU)
+        {
+            SetModel(behaviorName, model, inferenceDevice);
+        }
+
         /// <summary>
         /// Updates the Model for the agent. Any model currently assigned to the
         /// agent will be replaced with the provided one. If the arguments are
         /// <param name="model"> The model to use for inference.</param>
         /// <param name = "inferenceDevice"> Define on what device the model
         /// will be run.</param>
-        public void GiveModel(
+        public void SetModel(
-            m_PolicyFactory.GiveModel(behaviorName, model, inferenceDevice);
+            if (behaviorName == m_PolicyFactory.behaviorName &&
+                model == m_PolicyFactory.model &&
+                inferenceDevice == m_PolicyFactory.inferenceDevice)
+            {
+                // If everything is the same, don't make any changes.
+                return;
+            }
+
+            m_PolicyFactory.model = model;
+            m_PolicyFactory.inferenceDevice = inferenceDevice;
+            m_PolicyFactory.behaviorName = behaviorName;
+            ReloadPolicy();
+        }
+
+        internal void ReloadPolicy()
+        {
+            if (!m_Initialized)
+            {
+                // If we haven't initialized yet, no need to make any changes now; they'll
+                // happen in LazyInitialize later.
+                return;
+            }
             m_Brain?.Dispose();
             m_Brain = m_PolicyFactory.GeneratePolicy(Heuristic);
         }
             TimerStack.Instance.SetGauge(gaugeName, GetCumulativeReward());
         }
 
+        [Obsolete("Done() has been deprecated, use EndEpisode() instead.")]
+        public void Done()
+        {
+            EndEpisode();
+        }
+
-        public void Done()
+        public void EndEpisode()
         {
             NotifyAgentDone(DoneReason.DoneCalled);
             _AgentReset();
             // should stay the previous action before the Done(), so that it is properly recorded.
             if (m_Action.vectorActions == null)
             {
-                if (param.vectorActionSpaceType == SpaceType.Continuous)
-                {
-                    m_Action.vectorActions = new float[param.vectorActionSize[0]];
-                    m_Info.storedVectorActions = new float[param.vectorActionSize[0]];
-                }
-                else
-                {
-                    m_Action.vectorActions = new float[param.vectorActionSize.Length];
-                    m_Info.storedVectorActions = new float[param.vectorActionSize.Length];
-                }
+                m_Action.vectorActions = new float[param.numActions];
+                m_Info.storedVectorActions = new float[param.numActions];
+        [Obsolete("InitializeAgent() has been deprecated, use Initialize() instead.")]
+        public virtual void InitializeAgent()
+        {
+        }
+
         /// <summary>
         /// Initializes the agent, called once when the agent is enabled. Can be
         /// left empty if there is no special, unique set-up behavior for the
         /// One sample use is to store local references to other objects in the
         /// scene which would facilitate computing this agents observation.
         /// </remarks>
-        public virtual void InitializeAgent()
+        public virtual void Initialize()
+#pragma warning disable 0618
+            InitializeAgent();
+#pragma warning restore 0618
         }
 
         /// <summary>
         /// </returns>
         public virtual float[] Heuristic()
         {
-            throw new UnityAgentsException(
-                "The Heuristic method was not implemented for the Agent on the " +
-                $"{gameObject.name} GameObject.");
+            Debug.LogWarning("Heuristic method called but not implemented. Returning placeholder actions.");
+            var param = m_PolicyFactory.brainParameters;
+
+            return new float[param.numActions];
         }
 
         /// <summary>
         /// </summary>
         void SendInfoToBrain()
         {
+            if (!m_Initialized)
+            {
+                throw new UnityAgentsException("Call to SendInfoToBrain when Agent hasn't been initialized." +
+                    "Please ensure that you are calling 'base.OnEnable()' if you have overridden OnEnable.");
+            }
+
             if (m_Brain == null)
             {
                 return;
                     CollectDiscreteActionMasks(m_ActionMasker);
                 }
             }
-            m_Info.actionMasks = m_ActionMasker.GetMask();
+            m_Info.discreteActionMasks = m_ActionMasker.GetMask();
 
             m_Info.reward = m_Reward;
             m_Info.done = false;
             m_Brain.RequestDecision(m_Info, sensors);
 
             // If we have any DemonstrationWriters, write the AgentInfo and sensors to them.
-            foreach(var demoWriter in DemonstrationWriters)
+            foreach (var demoWriter in DemonstrationWriters)
             {
                 demoWriter.Record(m_Info, sensors);
             }
                 sensor.Update();
             }
         }
-
 
         /// <summary>
         /// Collects the vector observations of the agent.
         {
         }
 
+        [Obsolete("AgentAction() has been deprecated, use OnActionReceived() instead.")]
+        public virtual void AgentAction(float[] vectorAction)
+        {
+        }
+
         /// <summary>
         /// Specifies the agent behavior at every step based on the provided
         /// action.
         /// will be of length 1.
         /// </param>
-        public virtual void AgentAction(float[] vectorAction)
+        public virtual void OnActionReceived(float[] vectorAction)
+        {
+#pragma warning disable 0618
+            AgentAction(m_Action.vectorActions);
+#pragma warning restore 0618
+        }
+
+        [Obsolete("AgentReset() has been deprecated, use OnEpisodeBegin() instead.")]
+        public virtual void AgentReset()
         {
         }
 
         /// episode).
         /// </summary>
-        public virtual void AgentReset()
+        public virtual void OnEpisodeBegin()
+#pragma warning disable 0618
+            AgentReset();
+#pragma warning restore 0618
         }
 
         /// <summary>
         }
 
         /// <summary>
-        /// This method will forcefully reset the agent and will also reset the hasAlreadyReset flag.
-        /// This way, even if the agent was already in the process of reseting, it will be reset again
-        /// and will not send a Done flag at the next step.
-        /// </summary>
-        void ForceReset()
-        {
-            _AgentReset();
-        }
-
-        /// <summary>
         /// An internal reset method that updates internal data structures in
         /// addition to calling <see cref="AgentReset"/>.
         /// </summary>
             m_StepCount = 0;
-            AgentReset();
+            OnEpisodeBegin();
         }
 
         /// <summary>
             if ((m_RequestAction) && (m_Brain != null))
             {
                 m_RequestAction = false;
-                AgentAction(m_Action.vectorActions);
+                OnActionReceived(m_Action.vectorActions);
             }
 
             if ((m_StepCount >= maxStep) && (maxStep > 0))
         void DecideAction()
         {
             m_Action.vectorActions = m_Brain?.DecideAction();
-            if (m_Action.vectorActions == null){
+            if (m_Action.vectorActions == null)
+            {
                 ResetData();
             }
         }

com.unity.ml-agents/Runtime/Communicator/GrpcExtensions.cs

 using System.Runtime.CompilerServices;
 using MLAgents.Sensors;
 using MLAgents.Demonstrations;
+using MLAgents.Policies;
+
 
 [assembly: InternalsVisibleTo("Unity.ML-Agents.Editor")]
 [assembly: InternalsVisibleTo("Unity.ML-Agents.Editor.Tests")]
                 Id = ai.episodeId,
             };
 
-            if (ai.actionMasks != null)
+            if (ai.discreteActionMasks != null)
-                agentInfoProto.ActionMask.AddRange(ai.actionMasks);
+                agentInfoProto.ActionMask.AddRange(ai.discreteActionMasks);
             }
 
             return agentInfoProto;
         {
             return new UnityRLInitParameters
             {
-                seed = inputProto.Seed
+                seed = inputProto.Seed,
+                pythonLibraryVersion = inputProto.PackageVersion,
+                pythonCommunicationVersion = inputProto.CommunicationVersion,
             };
         }
 
             }
             else
             {
+                var compressedObs = sensor.GetCompressedObservation();
+                if (compressedObs == null)
+                {
+                    throw new UnityAgentsException(
+                        $"GetCompressedObservation() returned null data for sensor named {sensor.GetName()}. " +
+                        "You must return a byte[]. If you don't want to use compressed observations, " +
+                        "return SensorCompressionType.None from GetCompressionType()."
+                        );
+                }
+
-                    CompressedData = ByteString.CopyFrom(sensor.GetCompressedObservation()),
+                    CompressedData = ByteString.CopyFrom(compressedObs),
                     CompressionType = (CompressionTypeProto)sensor.GetCompressionType(),
                 };
             }

com.unity.ml-agents/Runtime/Communicator/ICommunicator.cs

 using System;
 using System.Collections.Generic;
-using UnityEngine;
+using MLAgents.Policies;
+using MLAgents.SideChannels;
 
 namespace MLAgents
 {
         /// Port to listen for connections on.
         /// </summary>
         public int port;
+
+
-        public string version;
+        public string unityPackageVersion;
+
+        /// <summary>
+        /// The version of the communication API.
+        /// </summary>
+        public string unityCommunicationVersion;
     }
     internal struct UnityRLInitParameters
     {
         public int seed;
+
+        /// <summary>
+        /// The library version of the python process.
+        /// </summary>
+        public string pythonLibraryVersion;
+
+        /// <summary>
+        /// The version of the communication API that python is using.
+        /// </summary>
+        public string pythonCommunicationVersion;
     }
     internal struct UnityRLInputParameters
     {

com.unity.ml-agents/Runtime/Communicator/RpcCommunicator.cs

 using UnityEngine;
 using MLAgents.CommunicatorObjects;
 using MLAgents.Sensors;
+using MLAgents.Policies;
+using MLAgents.SideChannels;
 using System.IO;
 using Google.Protobuf;
 
             var academyParameters = new UnityRLInitializationOutputProto
             {
                 Name = initParameters.name,
-                Version = initParameters.version
+                PackageVersion = initParameters.unityPackageVersion,
+                CommunicationVersion = initParameters.unityCommunicationVersion
             };
 
             UnityInputProto input;
                         RlInitializationOutput = academyParameters
                     },
                     out input);
+
+                // Initialization succeeded part-way. The most likely cause is a mismatch between the communicator
+                // API strings, so log an explicit warning if that's the case.
+                if (initializationInput != null && input == null)
+                {
+                    var pythonCommunicationVersion = initializationInput.RlInitializationInput.CommunicationVersion;
+                    var pythonPackageVersion = initializationInput.RlInitializationInput.PackageVersion;
+                    if (pythonCommunicationVersion != initParameters.unityCommunicationVersion)
+                    {
+                        Debug.LogWarningFormat(
+                            "Communication protocol between python ({0}) and Unity ({1}) don't match. " +
+                            "Python library version: {2}.",
+                            pythonCommunicationVersion, initParameters.unityCommunicationVersion,
+                            pythonPackageVersion
+                        );
+                    }
+                    else
+                    {
+                        Debug.LogWarningFormat(
+                            "Unknown communication error between Python. Python communication protocol: {0}, " +
+                            "Python library version: {1}.",
+                            pythonCommunicationVersion,
+                            pythonPackageVersion
+                        );
+                    }
+
+                    throw new UnityAgentsException("ICommunicator.Initialize() failed.");
+                }
             }
             catch
             {
                     "A side channel with type index {0} is already registered. You cannot register multiple " +
                     "side channels of the same id.", channelId));
             }
+
+            // Process any messages that we've already received for this channel ID.
+            var numMessages = m_CachedMessages.Count;
+            for (int i = 0; i < numMessages; i++)
+            {
+                var cachedMessage = m_CachedMessages.Dequeue();
+                if (channelId == cachedMessage.ChannelId)
+                {
+                    using (var incomingMsg = new IncomingMessage(cachedMessage.Message))
+                    {
+                        sideChannel.OnMessageReceived(incomingMsg);
+                    }
+                }
+                else
+                {
+                    m_CachedMessages.Enqueue(cachedMessage);
+                }
+            }
             m_SideChannels.Add(channelId, sideChannel);
         }
 
             }
         }
 
+        private struct CachedSideChannelMessage
+        {
+            public Guid ChannelId;
+            public byte[] Message;
+        }
+
+        private static Queue<CachedSideChannelMessage> m_CachedMessages = new Queue<CachedSideChannelMessage>();
+
         /// <summary>
         /// Separates the data received from Python into individual messages for each registered side channel.
         /// </summary>
         {
+            while (m_CachedMessages.Count != 0)
+            {
+                var cachedMessage = m_CachedMessages.Dequeue();
+                if (sideChannels.ContainsKey(cachedMessage.ChannelId))
+                {
+                    using (var incomingMsg = new IncomingMessage(cachedMessage.Message))
+                    {
+                        sideChannels[cachedMessage.ChannelId].OnMessageReceived(incomingMsg);
+                    }
+                }
+                else
+                {
+                    Debug.Log(string.Format(
+                        "Unknown side channel data received. Channel Id is "
+                        + ": {0}", cachedMessage.ChannelId));
+                }
+            }
+
             if (dataReceived.Length == 0)
             {
                 return;
                         }
                         if (sideChannels.ContainsKey(channelId))
                         {
-                            sideChannels[channelId].OnMessageReceived(message);
+                            using (var incomingMsg = new IncomingMessage(message))
+                            {
+                                sideChannels[channelId].OnMessageReceived(incomingMsg);
+                            }
-                            Debug.Log(string.Format(
-                                "Unknown side channel data received. Channel Id is "
-                                + ": {0}", channelId));
+                            // Don't recognize this ID, but cache it in case the SideChannel that can handle
+                            // it is registered before the next call to ProcessSideChannelData.
+                            m_CachedMessages.Enqueue(new CachedSideChannelMessage
+                            {
+                                ChannelId = channelId,
+                                Message = message
+                            });
                         }
                     }
                 }

com.unity.ml-agents/Runtime/DecisionRequester.cs

-using System.Runtime.CompilerServices;
+using UnityEngine.Serialization;
 
 namespace MLAgents
 {
     /// </summary>
     [AddComponentMenu("ML Agents/Decision Requester", (int)MenuGroup.Default)]
-    public class DecisionRequester : MonoBehaviour
+    internal class DecisionRequester : MonoBehaviour
     {
         /// <summary>
         /// The frequency with which the agent requests a decision. A DecisionPeriod of 5 means
         [Tooltip("Indicates whether or not the agent will take an action during the Academy " +
                  "steps where it does not request a decision. Has no effect when DecisionPeriod " +
                  "is set to 1.")]
-        public bool RepeatAction = true;
+        [FormerlySerializedAs("RepeatAction")]
+        public bool TakeActionsBetweenDecisions = true;
 
         /// <summary>
         /// Whether or not the Agent decisions should start at an offset (different for each agent).
             {
                 m_Agent?.RequestDecision();
             }
-            if (RepeatAction)
+            if (TakeActionsBetweenDecisions)
             {
                 m_Agent?.RequestAction();
             }

com.unity.ml-agents/Runtime/Demonstrations/Demonstration.cs

 using System;
 using UnityEngine;
+using MLAgents.Policies;
 
 namespace MLAgents.Demonstrations
 {

com.unity.ml-agents/Runtime/Demonstrations/DemonstrationRecorder.cs

 using System.Text.RegularExpressions;
 using UnityEngine;
 using System.IO;
+using MLAgents.Policies;
 
 namespace MLAgents.Demonstrations
 {
             var filePath = MakeDemonstrationFilePath(m_FileSystem, demonstrationDirectory, demonstrationName);
             var stream = m_FileSystem.File.Create(filePath);
             m_DemoWriter = new DemonstrationWriter(stream);
-
-            m_DemoWriter.Initialize(
-                demonstrationName,
-                behaviorParams.brainParameters,
-                behaviorParams.fullyQualifiedBehaviorName
-            );
 
             AddDemonstrationWriterToAgent(m_DemoWriter);
 
         /// <param name="demoWriter"></param>
         public void AddDemonstrationWriterToAgent(DemonstrationWriter demoWriter)
         {
+            var behaviorParams = GetComponent<BehaviorParameters>();
+            demoWriter.Initialize(
+                demonstrationName,
+                behaviorParams.brainParameters,
+                behaviorParams.fullyQualifiedBehaviorName
+            );
             m_Agent.DemonstrationWriters.Add(demoWriter);
         }
 

com.unity.ml-agents/Runtime/Demonstrations/DemonstrationWriter.cs

 using Google.Protobuf;
 using System.Collections.Generic;
 using MLAgents.Sensors;
+using MLAgents.Policies;
 
 namespace MLAgents.Demonstrations
 {
         /// <param name="demonstrationName">Base name of the demonstration file(s).</param>
         /// <param name="brainName">The name of the Brain the agent is attached to.</param>
         /// <param name="brainParameters">The parameters of the Brain the agent is attached to.</param>
-        public void Initialize(
+        internal void Initialize(
             string demonstrationName, BrainParameters brainParameters, string brainName)
         {
             if (m_Writer == null)

com.unity.ml-agents/Runtime/DiscreteActionMasker.cs

 using System;
 using System.Collections.Generic;
 using System.Linq;
+using MLAgents.Policies;
 
 namespace MLAgents
 {

com.unity.ml-agents/Runtime/Grpc/CommunicatorObjects/UnityRlInitializationInput.cs

           string.Concat(
             "CkZtbGFnZW50c19lbnZzL2NvbW11bmljYXRvcl9vYmplY3RzL3VuaXR5X3Js",
             "X2luaXRpYWxpemF0aW9uX2lucHV0LnByb3RvEhRjb21tdW5pY2F0b3Jfb2Jq",
-            "ZWN0cyIvCh9Vbml0eVJMSW5pdGlhbGl6YXRpb25JbnB1dFByb3RvEgwKBHNl",
-            "ZWQYASABKAVCH6oCHE1MQWdlbnRzLkNvbW11bmljYXRvck9iamVjdHNiBnBy",
-            "b3RvMw=="));
+            "ZWN0cyJnCh9Vbml0eVJMSW5pdGlhbGl6YXRpb25JbnB1dFByb3RvEgwKBHNl",
+            "ZWQYASABKAUSHQoVY29tbXVuaWNhdGlvbl92ZXJzaW9uGAIgASgJEhcKD3Bh",
+            "Y2thZ2VfdmVyc2lvbhgDIAEoCUIfqgIcTUxBZ2VudHMuQ29tbXVuaWNhdG9y",
+            "T2JqZWN0c2IGcHJvdG8z"));
-            new pbr::GeneratedClrTypeInfo(typeof(global::MLAgents.CommunicatorObjects.UnityRLInitializationInputProto), global::MLAgents.CommunicatorObjects.UnityRLInitializationInputProto.Parser, new[]{ "Seed" }, null, null, null)
+            new pbr::GeneratedClrTypeInfo(typeof(global::MLAgents.CommunicatorObjects.UnityRLInitializationInputProto), global::MLAgents.CommunicatorObjects.UnityRLInitializationInputProto.Parser, new[]{ "Seed", "CommunicationVersion", "PackageVersion" }, null, null, null)
           }));
     }
     #endregion
+  /// <summary>
+  /// The initializaiton message - this is typically sent from the Python trainer to the C# environment.
+  /// </summary>
   internal sealed partial class UnityRLInitializationInputProto : pb::IMessage<UnityRLInitializationInputProto> {
     private static readonly pb::MessageParser<UnityRLInitializationInputProto> _parser = new pb::MessageParser<UnityRLInitializationInputProto>(() => new UnityRLInitializationInputProto());
     private pb::UnknownFieldSet _unknownFields;
     [global::System.Diagnostics.DebuggerNonUserCodeAttribute]
     public UnityRLInitializationInputProto(UnityRLInitializationInputProto other) : this() {
       seed_ = other.seed_;
+      communicationVersion_ = other.communicationVersion_;
+      packageVersion_ = other.packageVersion_;
       _unknownFields = pb::UnknownFieldSet.Clone(other._unknownFields);
     }
 
       }
     }
 
+    /// <summary>Field number for the "communication_version" field.</summary>
+    public const int CommunicationVersionFieldNumber = 2;
+    private string communicationVersion_ = "";
+    /// <summary>
+    /// Communication protocol version that the initiating side (typically the Python trainer) is using.
+    /// </summary>
+    [global::System.Diagnostics.DebuggerNonUserCodeAttribute]
+    public string CommunicationVersion {
+      get { return communicationVersion_; }
+      set {
+        communicationVersion_ = pb::ProtoPreconditions.CheckNotNull(value, "value");
+      }
+    }
+
+    /// <summary>Field number for the "package_version" field.</summary>
+    public const int PackageVersionFieldNumber = 3;
+    private string packageVersion_ = "";
+    /// <summary>
+    /// Package/library version that the initiating side (typically the Python trainer) is using.
+    /// </summary>
+    [global::System.Diagnostics.DebuggerNonUserCodeAttribute]
+    public string PackageVersion {
+      get { return packageVersion_; }
+      set {
+        packageVersion_ = pb::ProtoPreconditions.CheckNotNull(value, "value");
+      }
+    }
+
     [global::System.Diagnostics.DebuggerNonUserCodeAttribute]
     public override bool Equals(object other) {
       return Equals(other as UnityRLInitializationInputProto);
         return true;
       }
       if (Seed != other.Seed) return false;
+      if (CommunicationVersion != other.CommunicationVersion) return false;
+      if (PackageVersion != other.PackageVersion) return false;
       return Equals(_unknownFields, other._unknownFields);
     }
 
       if (Seed != 0) hash ^= Seed.GetHashCode();
+      if (CommunicationVersion.Length != 0) hash ^= CommunicationVersion.GetHashCode();
+      if (PackageVersion.Length != 0) hash ^= PackageVersion.GetHashCode();
       if (_unknownFields != null) {
         hash ^= _unknownFields.GetHashCode();
       }
         output.WriteRawTag(8);
         output.WriteInt32(Seed);
       }
+      if (CommunicationVersion.Length != 0) {
+        output.WriteRawTag(18);
+        output.WriteString(CommunicationVersion);
+      }
+      if (PackageVersion.Length != 0) {
+        output.WriteRawTag(26);
+        output.WriteString(PackageVersion);
+      }
       if (_unknownFields != null) {
         _unknownFields.WriteTo(output);
       }
       if (Seed != 0) {
         size += 1 + pb::CodedOutputStream.ComputeInt32Size(Seed);
       }
+      if (CommunicationVersion.Length != 0) {
+        size += 1 + pb::CodedOutputStream.ComputeStringSize(CommunicationVersion);
+      }
+      if (PackageVersion.Length != 0) {
+        size += 1 + pb::CodedOutputStream.ComputeStringSize(PackageVersion);
+      }
       if (_unknownFields != null) {
         size += _unknownFields.CalculateSize();
       }
       if (other.Seed != 0) {
         Seed = other.Seed;
       }
+      if (other.CommunicationVersion.Length != 0) {
+        CommunicationVersion = other.CommunicationVersion;
+      }
+      if (other.PackageVersion.Length != 0) {
+        PackageVersion = other.PackageVersion;
+      }
       _unknownFields = pb::UnknownFieldSet.MergeFrom(_unknownFields, other._unknownFields);
     }
 
             break;
           case 8: {
             Seed = input.ReadInt32();
+            break;
+          }
+          case 18: {
+            CommunicationVersion = input.ReadString();
+            break;
+          }
+          case 26: {
+            PackageVersion = input.ReadString();
             break;
           }
         }

com.unity.ml-agents/Runtime/Grpc/CommunicatorObjects/UnityRlInitializationOutput.cs

             "CkdtbGFnZW50c19lbnZzL2NvbW11bmljYXRvcl9vYmplY3RzL3VuaXR5X3Js",
             "X2luaXRpYWxpemF0aW9uX291dHB1dC5wcm90bxIUY29tbXVuaWNhdG9yX29i",
             "amVjdHMaOW1sYWdlbnRzX2VudnMvY29tbXVuaWNhdG9yX29iamVjdHMvYnJh",
-            "aW5fcGFyYW1ldGVycy5wcm90byKfAQogVW5pdHlSTEluaXRpYWxpemF0aW9u",
-            "T3V0cHV0UHJvdG8SDAoEbmFtZRgBIAEoCRIPCgd2ZXJzaW9uGAIgASgJEhAK",
-            "CGxvZ19wYXRoGAMgASgJEkQKEGJyYWluX3BhcmFtZXRlcnMYBSADKAsyKi5j",
-            "b21tdW5pY2F0b3Jfb2JqZWN0cy5CcmFpblBhcmFtZXRlcnNQcm90b0oECAYQ",
-            "B0IfqgIcTUxBZ2VudHMuQ29tbXVuaWNhdG9yT2JqZWN0c2IGcHJvdG8z"));
+            "aW5fcGFyYW1ldGVycy5wcm90byLGAQogVW5pdHlSTEluaXRpYWxpemF0aW9u",
+            "T3V0cHV0UHJvdG8SDAoEbmFtZRgBIAEoCRIdChVjb21tdW5pY2F0aW9uX3Zl",
+            "cnNpb24YAiABKAkSEAoIbG9nX3BhdGgYAyABKAkSRAoQYnJhaW5fcGFyYW1l",
+            "dGVycxgFIAMoCzIqLmNvbW11bmljYXRvcl9vYmplY3RzLkJyYWluUGFyYW1l",
+            "dGVyc1Byb3RvEhcKD3BhY2thZ2VfdmVyc2lvbhgHIAEoCUoECAYQB0IfqgIc",
+            "TUxBZ2VudHMuQ29tbXVuaWNhdG9yT2JqZWN0c2IGcHJvdG8z"));
-            new pbr::GeneratedClrTypeInfo(typeof(global::MLAgents.CommunicatorObjects.UnityRLInitializationOutputProto), global::MLAgents.CommunicatorObjects.UnityRLInitializationOutputProto.Parser, new[]{ "Name", "Version", "LogPath", "BrainParameters" }, null, null, null)
+            new pbr::GeneratedClrTypeInfo(typeof(global::MLAgents.CommunicatorObjects.UnityRLInitializationOutputProto), global::MLAgents.CommunicatorObjects.UnityRLInitializationOutputProto.Parser, new[]{ "Name", "CommunicationVersion", "LogPath", "BrainParameters", "PackageVersion" }, null, null, null)
           }));
     }
     #endregion
     [global::System.Diagnostics.DebuggerNonUserCodeAttribute]
     public UnityRLInitializationOutputProto(UnityRLInitializationOutputProto other) : this() {
       name_ = other.name_;
-      version_ = other.version_;
+      communicationVersion_ = other.communicationVersion_;
+      packageVersion_ = other.packageVersion_;
       _unknownFields = pb::UnknownFieldSet.Clone(other._unknownFields);
     }
 
       }
     }
 
-    /// <summary>Field number for the "version" field.</summary>
-    public const int VersionFieldNumber = 2;
-    private string version_ = "";
+    /// <summary>Field number for the "communication_version" field.</summary>
+    public const int CommunicationVersionFieldNumber = 2;
+    private string communicationVersion_ = "";
+    /// <summary>
+    /// Communication protocol version that the responding side (typically the C# environment) is using.
+    /// </summary>
-    public string Version {
-      get { return version_; }
+    public string CommunicationVersion {
+      get { return communicationVersion_; }
-        version_ = pb::ProtoPreconditions.CheckNotNull(value, "value");
+        communicationVersion_ = pb::ProtoPreconditions.CheckNotNull(value, "value");
       }
     }
 
       get { return brainParameters_; }
     }
 
+    /// <summary>Field number for the "package_version" field.</summary>
+    public const int PackageVersionFieldNumber = 7;
+    private string packageVersion_ = "";
+    /// <summary>
+    /// Package/library version that the responding side (typically the C# environment) is using.
+    /// </summary>
+    [global::System.Diagnostics.DebuggerNonUserCodeAttribute]
+    public string PackageVersion {
+      get { return packageVersion_; }
+      set {
+        packageVersion_ = pb::ProtoPreconditions.CheckNotNull(value, "value");
+      }
+    }
+
     [global::System.Diagnostics.DebuggerNonUserCodeAttribute]
     public override bool Equals(object other) {
       return Equals(other as UnityRLInitializationOutputProto);
         return true;
       }
       if (Name != other.Name) return false;
-      if (Version != other.Version) return false;
+      if (CommunicationVersion != other.CommunicationVersion) return false;
+      if (PackageVersion != other.PackageVersion) return false;
       return Equals(_unknownFields, other._unknownFields);
     }
 
       if (Name.Length != 0) hash ^= Name.GetHashCode();
-      if (Version.Length != 0) hash ^= Version.GetHashCode();
+      if (CommunicationVersion.Length != 0) hash ^= CommunicationVersion.GetHashCode();
+      if (PackageVersion.Length != 0) hash ^= PackageVersion.GetHashCode();
       if (_unknownFields != null) {
         hash ^= _unknownFields.GetHashCode();
       }
         output.WriteRawTag(10);
         output.WriteString(Name);
       }
-      if (Version.Length != 0) {
+      if (CommunicationVersion.Length != 0) {
-        output.WriteString(Version);
+        output.WriteString(CommunicationVersion);
       }
       if (LogPath.Length != 0) {
         output.WriteRawTag(26);
+      if (PackageVersion.Length != 0) {
+        output.WriteRawTag(58);
+        output.WriteString(PackageVersion);
+      }
       if (_unknownFields != null) {
         _unknownFields.WriteTo(output);
       }
       if (Name.Length != 0) {
         size += 1 + pb::CodedOutputStream.ComputeStringSize(Name);
       }
-      if (Version.Length != 0) {
-        size += 1 + pb::CodedOutputStream.ComputeStringSize(Version);
+      if (CommunicationVersion.Length != 0) {
+        size += 1 + pb::CodedOutputStream.ComputeStringSize(CommunicationVersion);
+      if (PackageVersion.Length != 0) {
+        size += 1 + pb::CodedOutputStream.ComputeStringSize(PackageVersion);
+      }
       if (_unknownFields != null) {
         size += _unknownFields.CalculateSize();
       }
       if (other.Name.Length != 0) {
         Name = other.Name;
       }
-      if (other.Version.Length != 0) {
-        Version = other.Version;
+      if (other.CommunicationVersion.Length != 0) {
+        CommunicationVersion = other.CommunicationVersion;
+      if (other.PackageVersion.Length != 0) {
+        PackageVersion = other.PackageVersion;
+      }
       _unknownFields = pb::UnknownFieldSet.MergeFrom(_unknownFields, other._unknownFields);
     }
 
             break;
           }
           case 18: {
-            Version = input.ReadString();
+            CommunicationVersion = input.ReadString();
             break;
           }
           case 26: {
           case 42: {
             brainParameters_.AddEntriesFrom(input, _repeated_brainParameters_codec);
+            break;
+          }
+          case 58: {
+            PackageVersion = input.ReadString();
             break;
           }
         }

com.unity.ml-agents/Runtime/SideChannels/SideChannel.cs

 using System.Collections.Generic;
 using System;
+using System.IO;
+using System.Text;
-namespace MLAgents
+namespace MLAgents.SideChannels
 {
     /// <summary>
     /// Side channels provide an alternative mechanism of sending/receiving data from Unity
         /// of each type. Ensure the Unity side channels will be linked to their Python equivalent.
         /// </summary>
         /// <returns> The integer identifier of the SideChannel.</returns>
-        public Guid ChannelId{
+        public Guid ChannelId
+        {
             get;
             protected set;
         }
         /// Can be called multiple times per simulation step if multiple messages were sent.
         /// </summary>
-        /// <param name="data"> the payload of the message.</param>
-        public abstract void OnMessageReceived(byte[] data);
+        /// <param name="msg">The incoming message.</param>
+        public abstract void OnMessageReceived(IncomingMessage msg);
-        protected void QueueMessageToSend(byte[] data)
+        protected void QueueMessageToSend(OutgoingMessage msg)
-            MessageQueue.Add(data);
+            MessageQueue.Add(msg.ToByteArray());
         }
     }
 }

com.unity.ml-agents/Runtime/SideChannels/RawBytesChannel.cs

 using System.Collections.Generic;
 using System;
 
-namespace MLAgents
+namespace MLAgents.SideChannels
 {
     /// <summary>
     /// Side channel for managing raw bytes of data. It is up to the clients of this side channel
         }
 
         /// <inheritdoc/>
-        public override void OnMessageReceived(byte[] data)
+        public override void OnMessageReceived(IncomingMessage msg)
-            m_MessagesReceived.Add(data);
+            m_MessagesReceived.Add(msg.GetRawBytes());
         }
 
         /// <summary>
         /// <param name="data"> The byte array of data to send to Python.</param>
         public void SendRawBytes(byte[] data)
         {
-            QueueMessageToSend(data);
+            using (var msg = new OutgoingMessage())
+            {
+                msg.SetRawBytes(data);
+                QueueMessageToSend(msg);
+            }
         }
 
         /// <summary>

com.unity.ml-agents/Runtime/SideChannels/FloatPropertiesChannel.cs

 using System.Collections.Generic;
-using System.IO;
-using System.Text;
-namespace MLAgents
+namespace MLAgents.SideChannels
-    /// Interface for managing a collection of float properties keyed by a string variable.
-    /// </summary>
-    public interface IFloatProperties
-    {
-        /// <summary>
-        /// Sets one of the float properties of the environment. This data will be sent to Python.
-        /// </summary>
-        /// <param name="key"> The string identifier of the property.</param>
-        /// <param name="value"> The float value of the property.</param>
-        void SetProperty(string key, float value);
-
-        /// <summary>
-        /// Get an Environment property with a default value. If there is a value for this property,
-        /// it will be returned, otherwise, the default value will be returned.
-        /// </summary>
-        /// <param name="key"> The string identifier of the property.</param>
-        /// <param name="defaultValue"> The default value of the property.</param>
-        /// <returns></returns>
-        float GetPropertyWithDefault(string key, float defaultValue);
-
-        /// <summary>
-        /// Registers an action to be performed everytime the property is changed.
-        /// </summary>
-        /// <param name="key"> The string identifier of the property.</param>
-        /// <param name="action"> The action that ill be performed. Takes a float as input.</param>
-        void RegisterCallback(string key, Action<float> action);
-
-        /// <summary>
-        /// Returns a list of all the string identifiers of the properties currently present.
-        /// </summary>
-        /// <returns> The list of string identifiers </returns>
-        IList<string> ListProperties();
-    }
-
-    /// <summary>
-    public class FloatPropertiesChannel : SideChannel, IFloatProperties
+    public class FloatPropertiesChannel : SideChannel
     {
         Dictionary<string, float> m_FloatProperties = new Dictionary<string, float>();
         Dictionary<string, Action<float>> m_RegisteredActions = new Dictionary<string, Action<float>>();
             {
                 ChannelId = new Guid(k_FloatPropertiesDefaultId);
             }
-            else{
+            else
+            {
-        public override void OnMessageReceived(byte[] data)
+        public override void OnMessageReceived(IncomingMessage msg)
-            var kv = DeserializeMessage(data);
-            m_FloatProperties[kv.Key] = kv.Value;
-            if (m_RegisteredActions.ContainsKey(kv.Key))
-            {
-                m_RegisteredActions[kv.Key].Invoke(kv.Value);
-            }
+            var key = msg.ReadString();
+            var value = msg.ReadFloat32();
+
+            m_FloatProperties[key] = value;
+
+            Action<float> action;
+            m_RegisteredActions.TryGetValue(key, out action);
+            action?.Invoke(value);
         }
 
         /// <inheritdoc/>
-            QueueMessageToSend(SerializeMessage(key, value));
-            if (m_RegisteredActions.ContainsKey(key))
+            using (var msgOut = new OutgoingMessage())
-                m_RegisteredActions[key].Invoke(value);
+                msgOut.WriteString(key);
+                msgOut.WriteFloat32(value);
+                QueueMessageToSend(msgOut);
+
+            Action<float> action;
+            m_RegisteredActions.TryGetValue(key, out action);
+            action?.Invoke(value);
-            if (m_FloatProperties.ContainsKey(key))
-            {
-                return m_FloatProperties[key];
-            }
-            else
-            {
-                return defaultValue;
-            }
+            float valueOut;
+            bool hasKey = m_FloatProperties.TryGetValue(key, out valueOut);
+            return hasKey ? valueOut : defaultValue;
         }
 
         /// <inheritdoc/>
         public IList<string> ListProperties()
         {
             return new List<string>(m_FloatProperties.Keys);
-        }
-
-        static KeyValuePair<string, float> DeserializeMessage(byte[] data)
-        {
-            using (var memStream = new MemoryStream(data))
-            {
-                using (var binaryReader = new BinaryReader(memStream))
-                {
-                    var keyLength = binaryReader.ReadInt32();
-                    var key = Encoding.ASCII.GetString(binaryReader.ReadBytes(keyLength));
-                    var value = binaryReader.ReadSingle();
-                    return new KeyValuePair<string, float>(key, value);
-                }
-            }
-        }
-
-        static byte[] SerializeMessage(string key, float value)
-        {
-            using (var memStream = new MemoryStream())
-            {
-                using (var binaryWriter = new BinaryWriter(memStream))
-                {
-                    var stringEncoded = Encoding.ASCII.GetBytes(key);
-                    binaryWriter.Write(stringEncoded.Length);
-                    binaryWriter.Write(stringEncoded);
-                    binaryWriter.Write(value);
-                    return memStream.ToArray();
-                }
-            }
         }
     }
 }

com.unity.ml-agents/Runtime/SideChannels/EngineConfigurationChannel.cs

-using System.IO;
-namespace MLAgents
+namespace MLAgents.SideChannels
 {
     /// <summary>
     /// Side channel that supports modifying attributes specific to the Unity Engine.
         }
 
         /// <inheritdoc/>
-        public override void OnMessageReceived(byte[] data)
+        public override void OnMessageReceived(IncomingMessage msg)
-            using (var memStream = new MemoryStream(data))
-            {
-                using (var binaryReader = new BinaryReader(memStream))
-                {
-                    var width = binaryReader.ReadInt32();
-                    var height = binaryReader.ReadInt32();
-                    var qualityLevel = binaryReader.ReadInt32();
-                    var timeScale = binaryReader.ReadSingle();
-                    var targetFrameRate = binaryReader.ReadInt32();
+            var width = msg.ReadInt32();
+            var height = msg.ReadInt32();
+            var qualityLevel = msg.ReadInt32();
+            var timeScale = msg.ReadFloat32();
+            var targetFrameRate = msg.ReadInt32();
-                    timeScale = Mathf.Clamp(timeScale, 1, 100);
+            timeScale = Mathf.Clamp(timeScale, 1, 100);
-                    Screen.SetResolution(width, height, false);
-                    QualitySettings.SetQualityLevel(qualityLevel, true);
-                    Time.timeScale = timeScale;
-                    Time.captureFramerate = 60;
-                    Application.targetFrameRate = targetFrameRate;
-                }
-            }
+            Screen.SetResolution(width, height, false);
+            QualitySettings.SetQualityLevel(qualityLevel, true);
+            Time.timeScale = timeScale;
+            Time.captureFramerate = 60;
+            Application.targetFrameRate = targetFrameRate;
         }
     }
 }

com.unity.ml-agents/Runtime/Sensors/WriteAdapter.cs

 using System;
 using System.Collections.Generic;
 using Barracuda;
-using MLAgents.InferenceBrain;
+using MLAgents.Inference;
 
 namespace MLAgents.Sensors
 {

com.unity.ml-agents/Runtime/Sensors/RenderTextureSensorComponent.cs

 using UnityEngine;
+using UnityEngine.Serialization;
 
 namespace MLAgents.Sensors
 {
     [AddComponentMenu("ML Agents/Render Texture Sensor", (int)MenuGroup.Sensors)]
     public class RenderTextureSensorComponent : SensorComponent
     {
+        RenderTextureSensor m_Sensor;
+
-        public RenderTexture renderTexture;
+        [HideInInspector, SerializeField, FormerlySerializedAs("renderTexture")]
+        RenderTexture m_RenderTexture;
+
+        public RenderTexture renderTexture
+        {
+            get { return m_RenderTexture;  }
+            set { m_RenderTexture = value;  }
+        }
+
+        [HideInInspector, SerializeField, FormerlySerializedAs("sensorName")]
+        string m_SensorName = "RenderTextureSensor";
-        /// Name of the sensor.
+        /// Name of the generated <see cref="RenderTextureSensor"/>.
+        /// Note that changing this at runtime does not affect how the Agent sorts the sensors.
-        public string sensorName = "RenderTextureSensor";
+        public string sensorName
+        {
+            get { return m_SensorName;  }
+            set { m_SensorName = value; }
+        }
+
+        [HideInInspector, SerializeField, FormerlySerializedAs("grayscale")]
+        public bool m_Grayscale;
+        /// Note that changing this after the sensor is created has no effect.
-        public bool grayscale;
+        public bool grayscale
+        {
+            get { return m_Grayscale;  }
+            set { m_Grayscale = value; }
+        }
+
+        [HideInInspector, SerializeField, FormerlySerializedAs("compression")]
+        SensorCompressionType m_Compression = SensorCompressionType.PNG;
-        public SensorCompressionType compression = SensorCompressionType.PNG;
+        public SensorCompressionType compression
+        {
+            get { return m_Compression;  }
+            set { m_Compression = value; UpdateSensor(); }
+        }
-            return new RenderTextureSensor(renderTexture, grayscale, sensorName, compression);
+            m_Sensor = new RenderTextureSensor(renderTexture, grayscale, sensorName, compression);
+            return m_Sensor;
         }
 
         /// <inheritdoc/>
             var height = renderTexture != null ? renderTexture.height : 0;
 
             return new[] { height, width, grayscale ? 1 : 3 };
+        }
+
+        /// <summary>
+        /// Update fields that are safe to change on the Sensor at runtime.
+        /// </summary>
+        internal void UpdateSensor()
+        {
+            if (m_Sensor != null)
+            {
+                m_Sensor.compressionType = m_Compression;
+            }
         }
     }
 }

com.unity.ml-agents/Runtime/Sensors/RenderTextureSensor.cs

         SensorCompressionType m_CompressionType;
 
         /// <summary>
+        /// The compression type used by the sensor.
+        /// </summary>
+        public SensorCompressionType compressionType
+        {
+            get { return m_CompressionType;  }
+            set { m_CompressionType = value; }
+        }
+
+
+        /// <summary>
         /// Initializes the sensor.
         /// </summary>
         /// <param name="renderTexture">The <see cref="RenderTexture"/> instance to wrap.</param>
                 var texture = ObservationToTexture(m_RenderTexture);
                 // TODO support more types here, e.g. JPG
                 var compressed = texture.EncodeToPNG();
-                UnityEngine.Object.Destroy(texture);
+                DestroyTexture(texture);
                 return compressed;
             }
         }
             {
                 var texture = ObservationToTexture(m_RenderTexture);
                 var numWritten = Utilities.TextureToTensorProxy(texture, adapter, m_Grayscale);
-                UnityEngine.Object.Destroy(texture);
+                DestroyTexture(texture);
                 return numWritten;
             }
         }
             texture2D.Apply();
             RenderTexture.active = prevActiveRt;
             return texture2D;
+        }
+
+        static void DestroyTexture(Texture2D texture)
+        {
+            if (Application.isEditor)
+            {
+                // Edit Mode tests complain if we use Destroy()
+                // TODO move to extension methods for UnityEngine.Object?
+                UnityEngine.Object.DestroyImmediate(texture);
+            }
+            else
+            {
+                UnityEngine.Object.Destroy(texture);
+            }
         }
     }
 }

com.unity.ml-agents/Runtime/Sensors/RayPerceptionSensorComponentBase.cs

     /// </summary>
     public abstract class RayPerceptionSensorComponentBase : SensorComponent
     {
-        [HideInInspector]
-        [SerializeField]
-        [FormerlySerializedAs("sensorName")]
+        [HideInInspector, SerializeField, FormerlySerializedAs("sensorName")]
+        /// Note that changing this at runtime does not affect how the Agent sorts the sensors.
-            get => m_SensorName;
-            // Restrict the access on the name, since changing it a runtime doesn't re-sort the Agent sensors.
-            internal set => m_SensorName = value;
+            get { return m_SensorName; }
+            set { m_SensorName = value; }
-        [SerializeField]
-        [FormerlySerializedAs("detectableTags")]
+        [SerializeField, FormerlySerializedAs("detectableTags")]
+        /// Note that this should not be changed at runtime.
-            get => m_DetectableTags;
-            // Note: can't change at runtime
-            internal set => m_DetectableTags = value;
+            get { return m_DetectableTags; }
+            set { m_DetectableTags = value; }
-        [HideInInspector]
-        [SerializeField]
-        [FormerlySerializedAs("raysPerDirection")]
+        [HideInInspector, SerializeField, FormerlySerializedAs("raysPerDirection")]
         [Range(0, 50)]
         [Tooltip("Number of rays to the left and right of center.")]
         int m_RaysPerDirection = 3;
+        /// Note that this should not be changed at runtime.
-            get => m_RaysPerDirection;
+            get { return m_RaysPerDirection; }
-            internal set => m_RaysPerDirection = value;
+            set { m_RaysPerDirection = value;}
-        [HideInInspector]
-        [SerializeField]
-        [FormerlySerializedAs("maxRayDegrees")]
+        [HideInInspector, SerializeField, FormerlySerializedAs("maxRayDegrees")]
-                 "Greater than 90 degrees will go backwards.")]
+            "Greater than 90 degrees will go backwards.")]
         float m_MaxRayDegrees = 70;
 
         /// <summary>
             set { m_MaxRayDegrees = value; UpdateSensor(); }
         }
 
-        [HideInInspector]
-        [SerializeField]
-        [FormerlySerializedAs("sphereCastRadius")]
+        [HideInInspector, SerializeField, FormerlySerializedAs("sphereCastRadius")]
         [Range(0f, 10f)]
         [Tooltip("Radius of sphere to cast. Set to zero for raycasts.")]
         float m_SphereCastRadius = 0.5f;
             set { m_SphereCastRadius = value; UpdateSensor(); }
         }
 
-        [HideInInspector]
-        [SerializeField]
-        [FormerlySerializedAs("rayLength")]
+        [HideInInspector, SerializeField, FormerlySerializedAs("rayLength")]
         [Range(1, 1000)]
         [Tooltip("Length of the rays to cast.")]
         float m_RayLength = 20f;
             set { m_RayLength = value; UpdateSensor(); }
         }
 
-        [HideInInspector]
-        [SerializeField]
-        [FormerlySerializedAs("rayLayerMask")]
+        [HideInInspector, SerializeField, FormerlySerializedAs("rayLayerMask")]
         [Tooltip("Controls which layers the rays can hit.")]
         LayerMask m_RayLayerMask = Physics.DefaultRaycastLayers;
 
         public LayerMask rayLayerMask
         {
             get => m_RayLayerMask;
-            set { m_RayLayerMask = value; UpdateSensor();}
+            set { m_RayLayerMask = value; UpdateSensor(); }
-        [HideInInspector]
-        [SerializeField]
-        [FormerlySerializedAs("observationStacks")]
+        [HideInInspector, SerializeField, FormerlySerializedAs("observationStacks")]
         [Range(1, 50)]
         [Tooltip("Whether to stack previous observations. Using 1 means no previous observations.")]
         int m_ObservationStacks = 1;
+        /// Note that changing this after the sensor is created has no effect.
-        internal int observationStacks
+        public int observationStacks
-            get => m_ObservationStacks;
-            set => m_ObservationStacks = value; // Note: can't change at runtime
+            get { return m_ObservationStacks; }
+            set { m_ObservationStacks = value; }
         }
 
         /// <summary>
         RayPerceptionSensor m_RaySensor;
 
         /// <summary>
+        /// Get the RayPerceptionSensor that was created.
+        /// </summary>
+        public RayPerceptionSensor raySensor
+        {
+            get => m_RaySensor;
+        }
+
+        /// <summary>
         /// Returns the <see cref="RayPerceptionCastType"/> for the associated raycast sensor.
         /// </summary>
         /// <returns></returns>
             return new[] { obsSize * stacks };
         }
 
-        RayPerceptionInput GetRayPerceptionInput()
+        /// <summary>
+        /// Get the RayPerceptionInput that is used by the <see cref="RayPerceptionSensor"/>.
+        /// </summary>
+        /// <returns></returns>
+        public RayPerceptionInput GetRayPerceptionInput()
         {
             var rayAngles = GetRayAngles(raysPerDirection, maxRayDegrees);
 
         /// <summary>
         /// Draw the debug information from the sensor (if available).
         /// </summary>
-        void DrawRaycastGizmos(DebugDisplayInfo.RayInfo rayInfo, float alpha=1.0f)
+        void DrawRaycastGizmos(DebugDisplayInfo.RayInfo rayInfo, float alpha = 1.0f)
         {
             var startPositionWorld = rayInfo.worldStart;
             var endPositionWorld = rayInfo.worldEnd;

com.unity.ml-agents/Runtime/Sensors/RayPerceptionSensorComponent3D.cs

     [AddComponentMenu("ML Agents/Ray Perception Sensor 3D", (int)MenuGroup.Sensors)]
     public class RayPerceptionSensorComponent3D : RayPerceptionSensorComponentBase
     {
-        [HideInInspector]
-        [SerializeField]
-        [FormerlySerializedAs("startVerticalOffset")]
+        [HideInInspector, SerializeField, FormerlySerializedAs("startVerticalOffset")]
         [Range(-10f, 10f)]
         [Tooltip("Ray start is offset up or down by this amount.")]
         float m_StartVerticalOffset;
             set { m_StartVerticalOffset = value; UpdateSensor(); }
         }
 
-        [HideInInspector]
-        [SerializeField]
-        [FormerlySerializedAs("endVerticalOffset")]
+        [HideInInspector, SerializeField, FormerlySerializedAs("endVerticalOffset")]
         [Range(-10f, 10f)]
         [Tooltip("Ray end is offset up or down by this amount.")]
         float m_EndVerticalOffset;

com.unity.ml-agents/Runtime/Sensors/RayPerceptionSensor.cs

             var startPositionWorld = transform.TransformPoint(startPositionLocal);
             var endPositionWorld = transform.TransformPoint(endPositionLocal);
 
-            return (StartPositionWorld: startPositionWorld, EndPositionWorld: endPositionWorld);
+            return (StartPositionWorld : startPositionWorld, EndPositionWorld : endPositionWorld);
         }
 
         /// <summary>
         /// <param name="rayInput">The inputs for the sensor.</param>
         public RayPerceptionSensor(string name, RayPerceptionInput rayInput)
         {
-            var numObservations = rayInput.OutputSize();
-            m_Shape = new[] { numObservations };
-            m_Observations = new float[numObservations];
+            SetNumObservations(rayInput.OutputSize());
 
             if (Application.isEditor)
             {
 
-        internal void SetRayPerceptionInput(RayPerceptionInput input)
+        void SetNumObservations(int numObservations)
-            // TODO make sure that number of rays and tags don't change
-            m_RayPerceptionInput = input;
+            m_Shape = new[] { numObservations };
+            m_Observations = new float[numObservations];
+        }
+
+        internal void SetRayPerceptionInput(RayPerceptionInput rayInput)
+        {
+            // Note that change the number of rays or tags doesn't directly call this,
+            // but changing them and then changing another field will.
+            if (m_RayPerceptionInput.OutputSize() != rayInput.OutputSize())
+            {
+                Debug.Log(
+                    "Changing the number of tags or rays at runtime is not " +
+                    "supported and may cause errors in training or inference."
+                );
+                // Changing the shape will probably break things downstream, but we can at least
+                // keep this consistent.
+                SetNumObservations(rayInput.OutputSize());
+            }
+            m_RayPerceptionInput = rayInput;
         }
 
         /// <summary>

com.unity.ml-agents/Runtime/Sensors/CameraSensorComponent.cs

 using UnityEngine;
+using UnityEngine.Serialization;
 
 namespace MLAgents.Sensors
 {
     [AddComponentMenu("ML Agents/Camera Sensor", (int)MenuGroup.Sensors)]
     public class CameraSensorComponent : SensorComponent
     {
+        [HideInInspector, SerializeField, FormerlySerializedAs("camera")]
+        Camera m_Camera;
+
+        CameraSensor m_Sensor;
+
-        public new Camera camera;
+        public new Camera camera
+        {
+            get { return m_Camera;  }
+            set { m_Camera = value; UpdateSensor(); }
+        }
+
+        [HideInInspector, SerializeField, FormerlySerializedAs("sensorName")]
+        string m_SensorName = "CameraSensor";
+        /// Note that changing this at runtime does not affect how the Agent sorts the sensors.
-        public string sensorName = "CameraSensor";
+        public string sensorName
+        {
+            get { return m_SensorName;  }
+            set { m_SensorName = value; }
+        }
+
+        [HideInInspector, SerializeField, FormerlySerializedAs("width")]
+        int m_Width = 84;
-        /// Width of the generated image.
+        /// Width of the generated observation.
+        /// Note that changing this after the sensor is created has no effect.
-        public int width = 84;
+        public int width
+        {
+            get { return m_Width;  }
+            set { m_Width = value; }
+        }
+
+        [HideInInspector, SerializeField, FormerlySerializedAs("height")]
+        int m_Height = 84;
-        /// Height of the generated image.
+        /// Height of the generated observation.
+        /// Note that changing this after the sensor is created has no effect.
-        public int height = 84;
+        public int height
+        {
+            get { return m_Height;  }
+            set { m_Height = value;  }
+        }
+
+        [HideInInspector, SerializeField, FormerlySerializedAs("grayscale")]
+        public bool m_Grayscale;
+        /// Note that changing this after the sensor is created has no effect.
-        public bool grayscale;
+        public bool grayscale
+        {
+            get { return m_Grayscale;  }
+            set { m_Grayscale = value; }
+        }
+
+        [HideInInspector, SerializeField, FormerlySerializedAs("compression")]
+        SensorCompressionType m_Compression = SensorCompressionType.PNG;
-        public SensorCompressionType compression = SensorCompressionType.PNG;
+        public SensorCompressionType compression
+        {
+            get { return m_Compression;  }
+            set { m_Compression = value; UpdateSensor(); }
+        }
 
         /// <summary>
         /// Creates the <see cref="CameraSensor"/>
         {
-            return new CameraSensor(camera, width, height, grayscale, sensorName, compression);
+            m_Sensor = new CameraSensor(m_Camera, m_Width, m_Height, grayscale, m_SensorName, compression);
+            return m_Sensor;
         }
 
         /// <summary>
         public override int[] GetObservationShape()
         {
-            return CameraSensor.GenerateShape(width, height, grayscale);
+            return CameraSensor.GenerateShape(m_Width, m_Height, grayscale);
+        }
+
+        /// <summary>
+        /// Update fields that are safe to change on the Sensor at runtime.
+        /// </summary>
+        internal void UpdateSensor()
+        {
+            if (m_Sensor != null)
+            {
+                m_Sensor.camera = m_Camera;
+                m_Sensor.compressionType = m_Compression;
+            }
         }
     }
 }

com.unity.ml-agents/Runtime/Sensors/CameraSensor.cs

         SensorCompressionType m_CompressionType;
 
         /// <summary>
+        /// The Camera used for rendering the sensor observations.
+        /// </summary>
+        public Camera camera
+        {
+            get { return m_Camera; }
+            set { m_Camera = value; }
+        }
+
+        /// <summary>
+        /// The compression type used by the sensor.
+        /// </summary>
+        public SensorCompressionType compressionType
+        {
+            get { return m_CompressionType;  }
+            set { m_CompressionType = value; }
+        }
+
+
+        /// <summary>
         /// Creates and returns the camera sensor.
         /// </summary>
         /// <param name="camera">Camera object to capture images from.</param>

com.unity.ml-agents/Runtime/Policies/RemotePolicy.cs

 using System;
 using MLAgents.Sensors;
 
-namespace MLAgents
+namespace MLAgents.Policies
 {
     /// <summary>
     /// The Remote Policy only works when training.

com.unity.ml-agents/Runtime/Policies/IPolicy.cs

 using System.Collections.Generic;
 using MLAgents.Sensors;
 
-namespace MLAgents
+namespace MLAgents.Policies
 {
     /// <summary>
     /// IPolicy is connected to a single Agent. Each time the agent needs

com.unity.ml-agents/Runtime/Policies/HeuristicPolicy.cs

 using System.Collections.Generic;
 using System;
+using System.Collections;
-namespace MLAgents
+namespace MLAgents.Policies
 {
     /// <summary>
     /// The Heuristic Policy uses a hards coded Heuristic method
         Func<float[]> m_Heuristic;
         float[] m_LastDecision;
 
+        WriteAdapter m_WriteAdapter = new WriteAdapter();
+        NullList m_NullList = new NullList();
+
+
         /// <inheritdoc />
         public HeuristicPolicy(Func<float[]> heuristic)
         {
         /// <inheritdoc />
         public void RequestDecision(AgentInfo info, List<ISensor> sensors)
         {
+            StepSensors(sensors);
             m_LastDecision = m_Heuristic.Invoke();
         }
 
 
         public void Dispose()
         {
+        }
+
+        /// <summary>
+        /// Trivial implementation of the IList interface that does nothing.
+        /// This is only used for "writing" observations that we will discard.
+        /// </summary>
+        class NullList : IList<float>
+        {
+            public IEnumerator<float> GetEnumerator()
+            {
+                throw new NotImplementedException();
+            }
+
+            IEnumerator IEnumerable.GetEnumerator()
+            {
+                return GetEnumerator();
+            }
+
+            public void Add(float item)
+            {
+            }
+
+            public void Clear()
+            {
+            }
+
+            public bool Contains(float item)
+            {
+                return false;
+            }
+
+            public void CopyTo(float[] array, int arrayIndex)
+            {
+                throw new NotImplementedException();
+            }
+
+            public bool Remove(float item)
+            {
+                return false;
+            }
+
+            public int Count { get; }
+            public bool IsReadOnly { get; }
+            public int IndexOf(float item)
+            {
+                return -1;
+            }
+
+            public void Insert(int index, float item)
+            {
+            }
+
+            public void RemoveAt(int index)
+            {
+            }
+
+            public float this[int index]
+            {
+                get { return 0.0f; }
+                set { }
+            }
+        }
+
+        /// <summary>
+        /// Run ISensor.Write or ISensor.GetCompressedObservation for each sensor
+        /// The output is currently unused, but this makes the sensor usage consistent
+        /// between training and inference.
+        /// </summary>
+        /// <param name="sensors"></param>
+        void StepSensors(List<ISensor> sensors)
+        {
+            foreach (var sensor in sensors)
+            {
+                if (sensor.GetCompressionType() == SensorCompressionType.None)
+                {
+                    m_WriteAdapter.SetTarget(m_NullList, sensor.GetObservationShape(), 0);
+                    sensor.Write(m_WriteAdapter);
+                }
+                else
+                {
+                    sensor.GetCompressedObservation();
+                }
+            }
         }
     }
 }

com.unity.ml-agents/Runtime/Policies/BrainParameters.cs

 using System;
 using UnityEngine;
 
-namespace MLAgents
+namespace MLAgents.Policies
 {
     /// <summary>
     /// Whether the action space is discrete or continuous.
         /// Defines if the action is discrete or continuous.
         /// </summary>
         public SpaceType vectorActionSpaceType = SpaceType.Discrete;
+
+        public int numActions
+        {
+            get
+            {
+                switch (vectorActionSpaceType)
+                {
+                    case SpaceType.Discrete:
+                        return vectorActionSize.Length;
+                    case SpaceType.Continuous:
+                        return vectorActionSize[0];
+                    default:
+                        return 0;
+                }
+            }
+        }
 
         /// <summary>
         /// Deep clones the BrainParameter object.

com.unity.ml-agents/Runtime/Policies/BehaviorParameters.cs

 using UnityEngine;
 using UnityEngine.Serialization;
 
-namespace MLAgents
+namespace MLAgents.Policies
+    /// Defines what type of behavior the Agent will be using
+    /// </summary>
+    [Serializable]
+    public enum BehaviorType
+    {
+        /// <summary>
+        /// The Agent will use the remote process for decision making.
+        /// if unavailable, will use inference and if no model is provided, will use
+        /// the heuristic.
+        /// </summary>
+        Default,
+
+        /// <summary>
+        /// The Agent will always use its heuristic
+        /// </summary>
+        HeuristicOnly,
+
+        /// <summary>
+        /// The Agent will always use inference with the provided
+        /// neural network model.
+        /// </summary>
+        InferenceOnly
+    }
+
+
+    /// <summary>
-        [Serializable]
-        enum BehaviorType
+        [HideInInspector, SerializeField]
+        BrainParameters m_BrainParameters = new BrainParameters();
+
+        /// <summary>
+        /// The associated <see cref="BrainParameters"/> for this behavior.
+        /// </summary>
+        public BrainParameters brainParameters
-            Default,
-            HeuristicOnly,
-            InferenceOnly
+            get { return m_BrainParameters; }
+            internal set { m_BrainParameters = value; }
-        [HideInInspector]
-        [SerializeField]
-        BrainParameters m_BrainParameters = new BrainParameters();
-        [HideInInspector]
-        [SerializeField]
+        [HideInInspector, SerializeField]
-        [HideInInspector]
-        [SerializeField]
-        InferenceDevice m_InferenceDevice;
-        [HideInInspector]
-        [SerializeField]
-
-        // Disable warning /com.unity.ml-agents/Runtime/Policy/BehaviorParameters.cs(...):
-        //   warning CS0649: Field 'BehaviorParameters.m_BehaviorType' is never assigned to,
-        //   and will always have its default value
-        // This field is set in the custom editor.
-        #pragma warning disable 0649
-        BehaviorType m_BehaviorType;
-        #pragma warning restore 0649
-        [HideInInspector]
-        [SerializeField]
-        string m_BehaviorName = "My Behavior";
-        /// The team ID for this behavior.
+        /// The neural network model used when in inference mode.
+        /// This should not be set at runtime; use <see cref="Agent.SetModel(string,NNModel,InferenceDevice)"/>
+        /// to set it instead.
-        [HideInInspector]
-        [SerializeField]
-        public int m_TeamID;
-        [FormerlySerializedAs("m_useChildSensors")]
-        [HideInInspector]
-        [SerializeField]
-        [Tooltip("Use all Sensor components attached to child GameObjects of this Agent.")]
-        bool m_UseChildSensors = true;
+        public NNModel model
+        {
+            get { return m_Model; }
+            set { m_Model = value; UpdateAgentPolicy(); }
+        }
+
+        [HideInInspector, SerializeField]
+        InferenceDevice m_InferenceDevice;
-        /// The associated <see cref="BrainParameters"/> for this behavior.
+        /// How inference is performed for this Agent's model.
+        /// This should not be set at runtime; use <see cref="Agent.SetModel(string,NNModel,InferenceDevice)"/>
+        /// to set it instead.
-        public BrainParameters brainParameters
+        public InferenceDevice inferenceDevice
-            get { return m_BrainParameters; }
+            get { return m_InferenceDevice; }
+            set { m_InferenceDevice = value; UpdateAgentPolicy();}
+        [HideInInspector, SerializeField]
+        BehaviorType m_BehaviorType;
+
-        /// Whether or not to use all the sensor components attached to child GameObjects of the agent.
+        /// The BehaviorType for the Agent.
-        public bool useChildSensors
+        public BehaviorType behaviorType
-            get { return m_UseChildSensors; }
+            get { return m_BehaviorType; }
+            set { m_BehaviorType = value; UpdateAgentPolicy(); }
+        [HideInInspector, SerializeField]
+        string m_BehaviorName = "My Behavior";
+
+        /// This should not be set at runtime; use <see cref="Agent.SetModel(string,NNModel,InferenceDevice)"/>
+        /// to set it instead.
+            set { m_BehaviorName = value; UpdateAgentPolicy(); }
+        }
+
+        /// <summary>
+        /// The team ID for this behavior.
+        /// </summary>
+        [HideInInspector, SerializeField, FormerlySerializedAs("m_TeamID")]
+        public int TeamId;
+        // TODO properties here instead of Agent
+
+        [FormerlySerializedAs("m_useChildSensors")]
+        [HideInInspector]
+        [SerializeField]
+        [Tooltip("Use all Sensor components attached to child GameObjects of this Agent.")]
+        bool m_UseChildSensors = true;
+
+        /// <summary>
+        /// Whether or not to use all the sensor components attached to child GameObjects of the agent.
+        /// Note that changing this after the Agent has been initialized will not have any effect.
+        /// </summary>
+        public bool useChildSensors
+        {
+            get { return m_UseChildSensors; }
+            set { m_UseChildSensors = value; }
         }
 
         /// <summary>
         {
-            get { return m_BehaviorName + "?team=" + m_TeamID; }
+            get { return m_BehaviorName + "?team=" + TeamId; }
         }
 
         internal IPolicy GeneratePolicy(Func<float[]> heuristic)
                 case BehaviorType.HeuristicOnly:
                     return new HeuristicPolicy(heuristic);
                 case BehaviorType.InferenceOnly:
+                {
+                    if (m_Model == null)
+                    {
+                        var behaviorType = BehaviorType.InferenceOnly.ToString();
+                        throw new UnityAgentsException(
+                            $"Can't use Behavior Type {behaviorType} without a model. " +
+                            "Either assign a model, or change to a different Behavior Type."
+                        );
+                    }
+                }
                 case BehaviorType.Default:
                     if (Academy.Instance.IsCommunicatorOn)
                     {
             }
         }
 
-        /// <summary>
-        /// Updates the model and related details for this behavior.
-        /// </summary>
-        /// <param name="newBehaviorName">New name for the behavior.</param>
-        /// <param name="model">New neural network model for this behavior.</param>
-        /// <param name="inferenceDevice">New inference device for this behavior.</param>
-        public void GiveModel(
-            string newBehaviorName,
-            NNModel model,
-            InferenceDevice inferenceDevice = InferenceDevice.CPU)
+        internal void UpdateAgentPolicy()
-            m_Model = model;
-            m_InferenceDevice = inferenceDevice;
-            m_BehaviorName = newBehaviorName;
+            var agent = GetComponent<Agent>();
+            if (agent == null)
+            {
+                return;
+            }
+            agent.ReloadPolicy();
         }
     }
 }

com.unity.ml-agents/Runtime/Policies/BarracudaPolicy.cs

-using UnityEngine;
-using MLAgents.InferenceBrain;
-using System;
+using MLAgents.Inference;
-namespace MLAgents
+namespace MLAgents.Policies
 {
     /// <summary>
     /// Where to perform inference.

com.unity.ml-agents/Runtime/Inference/Utils/RandomNormal.cs

 using System;
 
-namespace MLAgents.InferenceBrain.Utils
+namespace MLAgents.Inference.Utils
 {
     /// <summary>
     /// RandomNormal - A random number generator that produces normally distributed random

com.unity.ml-agents/Runtime/Inference/Utils/Multinomial.cs

-namespace MLAgents.InferenceBrain.Utils
+namespace MLAgents.Inference.Utils
 {
     /// <summary>
     /// Multinomial - Draws samples from a multinomial distribution given a (potentially unscaled)

com.unity.ml-agents/Runtime/Inference/TensorProxy.cs

 using System;
 using System.Collections.Generic;
 using Barracuda;
-using MLAgents.InferenceBrain.Utils;
+using MLAgents.Inference.Utils;
-namespace MLAgents.InferenceBrain
+namespace MLAgents.Inference
 {
     /// <summary>
     /// Tensor - A class to encapsulate a Tensor used for inference.

com.unity.ml-agents/Runtime/Inference/TensorNames.cs

-namespace MLAgents.InferenceBrain
+namespace MLAgents.Inference
 {
     /// <summary>
     /// Contains the names of the input and output tensors for the Inference Brain.

com.unity.ml-agents/Runtime/Inference/TensorGenerator.cs

 using Barracuda;
 using MLAgents.Sensors;
 
-namespace MLAgents.InferenceBrain
+namespace MLAgents.Inference
 {
     /// <summary>
     /// Mapping between Tensor names and generators.

com.unity.ml-agents/Runtime/Inference/TensorApplier.cs

 using System.Collections.Generic;
 using Barracuda;
-using System;
+using MLAgents.Policies;
-namespace MLAgents.InferenceBrain
+namespace MLAgents.Inference
 {
     /// <summary>
     /// Mapping between the output tensor names and the method that will use the

com.unity.ml-agents/Runtime/Inference/ModelRunner.cs

 using Barracuda;
 using UnityEngine.Profiling;
 using MLAgents.Sensors;
+using MLAgents.Policies;
-namespace MLAgents.InferenceBrain
+namespace MLAgents.Inference
 {
     internal struct AgentInfoSensorsPair
     {
                 m_VisualObservationsInitialized = true;
             }
 
-            Profiler.BeginSample("LearningBrain.DecideAction");
+            Profiler.BeginSample("ModelRunner.DecideAction");
 
             Profiler.BeginSample($"MLAgents.{m_Model.name}.GenerateTensors");
             // Prepare the input tensors to be feed into the engine

com.unity.ml-agents/Runtime/Inference/GeneratorImpl.cs

 using System.Collections.Generic;
 using System;
 using Barracuda;
-using MLAgents.InferenceBrain.Utils;
+using MLAgents.Inference.Utils;
-namespace MLAgents.InferenceBrain
+namespace MLAgents.Inference
 {
     /// <summary>
     /// Reshapes a Tensor so that its first dimension becomes equal to the current batch size
             foreach (var infoSensorPair in infos)
             {
                 var agentInfo = infoSensorPair.agentInfo;
-                var maskList = agentInfo.actionMasks;
+                var maskList = agentInfo.discreteActionMasks;
                 for (var j = 0; j < maskSize; j++)
                 {
                     var isUnmasked = (maskList != null && maskList[j]) ? 0.0f : 1.0f;

com.unity.ml-agents/Runtime/Inference/BarracudaModelParamLoader.cs

 using System.Linq;
 using Barracuda;
 using MLAgents.Sensors;
+using MLAgents.Policies;
-namespace MLAgents.InferenceBrain
+namespace MLAgents.Inference
 {
     /// <summary>
     /// Prepares the Tensors for the Learning Brain and exposes a list of failed checks if Model
         /// </param>
         /// <param name="sensorComponents">Attached sensor components</param>
         /// <returns>The list the error messages of the checks that failed</returns>
-        public static IEnumerable<string> CheckModel(Model model, BrainParameters brainParameters, SensorComponent[] sensorComponents)
+        public static IEnumerable<string> CheckModel(Model model, BrainParameters brainParameters,
+            SensorComponent[] sensorComponents, BehaviorType behaviorType = BehaviorType.Default)
-                failedModelChecks.Add(
-                    "There is no model for this Brain, cannot run inference. " +
-                    "(But can still train)");
+                var errorMsg = "There is no model for this Brain; cannot run inference. ";
+                if (behaviorType == BehaviorType.InferenceOnly)
+                {
+                    errorMsg += "Either assign a model, or change to a different Behavior Type.";
+                }
+                else
+                {
+                    errorMsg += "(But can still train)";
+                }
+                failedModelChecks.Add(errorMsg);
                 return failedModelChecks;
             }
 

com.unity.ml-agents/Runtime/Inference/ApplierImpl.cs

 using System.Collections.Generic;
 using System.Linq;
 using Barracuda;
-using MLAgents.InferenceBrain.Utils;
+using MLAgents.Inference.Utils;
-namespace MLAgents.InferenceBrain
+namespace MLAgents.Inference
 {
     /// <summary>
     /// The Applier for the Continuous Action output tensor. Tensor is assumed to contain the

com.unity.ml-agents/Tests/Editor/DemonstrationTests.cs

 using MLAgents.CommunicatorObjects;
 using MLAgents.Sensors;
 using MLAgents.Demonstrations;
+using MLAgents.Policies;
 
 namespace MLAgents.Tests
 {
             var agentInfo = new AgentInfo
             {
                 reward = 1f,
-                actionMasks = new[] { false, true },
+                discreteActionMasks = new[] { false, true },
                 done = true,
                 episodeId = 5,
                 maxStepReached = true,

com.unity.ml-agents/Tests/Editor/DiscreteActionOutputApplierTest.cs

 using Barracuda;
 using NUnit.Framework;
 using UnityEngine;
-using MLAgents.InferenceBrain;
-using MLAgents.InferenceBrain.Utils;
+using MLAgents.Inference;
+using MLAgents.Inference.Utils;
 
 namespace MLAgents.Tests
 {

com.unity.ml-agents/Tests/Editor/EditModeTestActionMasker.cs

 using NUnit.Framework;
+using MLAgents.Policies;
 
 namespace MLAgents.Tests
 {

com.unity.ml-agents/Tests/Editor/EditModeTestInternalBrainTensorApplier.cs

 using System.Collections.Generic;
 using NUnit.Framework;
-using UnityEngine;
-using MLAgents.InferenceBrain;
-using System;
+using MLAgents.Inference;
+using MLAgents.Policies;
 
 namespace MLAgents.Tests
 {

com.unity.ml-agents/Tests/Editor/EditModeTestInternalBrainTensorGenerator.cs

 using Barracuda;
 using NUnit.Framework;
 using UnityEngine;
-using MLAgents.InferenceBrain;
-
+using MLAgents.Inference;
+using MLAgents.Policies;
 
 namespace MLAgents.Tests
 {
             var infoA = new AgentInfo
             {
                 storedVectorActions = new[] { 1f, 2f },
-                actionMasks = null
+                discreteActionMasks = null
-                actionMasks = new[] { true, false, false, false, false },
+                discreteActionMasks = new[] { true, false, false, false, false },
             };
 
 

com.unity.ml-agents/Tests/Editor/MLAgentsEditModeTest.cs

+using System.CodeDom;
+using MLAgents.Policies;
 
 namespace MLAgents.Tests
 {
             typeof(Agent).GetField("m_Brain", BindingFlags.Instance | BindingFlags.NonPublic).SetValue(this, policy);
         }
 
+        internal IPolicy GetPolicy()
+        {
+            return (IPolicy)typeof(Agent).GetField("m_Brain", BindingFlags.Instance | BindingFlags.NonPublic).GetValue(this);
+        }
+
-        public int collectObservationsCallsSinceLastReset;
+        public int collectObservationsCallsForEpisode;
-        public int agentActionCallsSinceLastReset;
-        public int agentResetCalls;
-        public override void InitializeAgent()
+        public int agentActionCallsForEpisode;
+        public int agentOnEpisodeBeginCalls;
+        public int heuristicCalls;
+        public TestSensor sensor1;
+        public TestSensor sensor2;
+
+        public override void Initialize()
-            var sensor1 = new TestSensor("testsensor1");
-            var sensor2 = new TestSensor("testsensor2");
+            sensor1 = new TestSensor("testsensor1");
+            sensor2 = new TestSensor("testsensor2");
+            sensor2.compressionType = SensorCompressionType.PNG;
 
             sensors.Add(sensor2);
             sensors.Add(sensor1);
         {
             collectObservationsCalls += 1;
-            collectObservationsCallsSinceLastReset += 1;
+            collectObservationsCallsForEpisode += 1;
-        public override void AgentAction(float[] vectorAction)
+        public override void OnActionReceived(float[] vectorAction)
-            agentActionCallsSinceLastReset += 1;
+            agentActionCallsForEpisode += 1;
-        public override void AgentReset()
+        public override void OnEpisodeBegin()
-            agentResetCalls += 1;
-            collectObservationsCallsSinceLastReset = 0;
-            agentActionCallsSinceLastReset = 0;
+            agentOnEpisodeBeginCalls += 1;
+            collectObservationsCallsForEpisode = 0;
+            agentActionCallsForEpisode = 0;
+            heuristicCalls++;
             return new float[0];
         }
     }
         public string sensorName;
+        public int numWriteCalls;
+        public int numCompressedCalls;
+        public SensorCompressionType compressionType = SensorCompressionType.None;
 
         public TestSensor(string n)
         {
 
         public int Write(WriteAdapter adapter)
         {
+            numWriteCalls++;
             // No-op
             return 0;
         }
-            return null;
+            numCompressedCalls++;
+            return new byte[] { 0 };
-            return SensorCompressionType.None;
+            return compressionType;
         }
 
         public string GetName()
             agentGo2.AddComponent<TestAgent>();
             var agent2 = agentGo2.GetComponent<TestAgent>();
 
-            Assert.AreEqual(0, agent1.agentResetCalls);
-            Assert.AreEqual(0, agent2.agentResetCalls);
+            Assert.AreEqual(0, agent1.agentOnEpisodeBeginCalls);
+            Assert.AreEqual(0, agent2.agentOnEpisodeBeginCalls);
             Assert.AreEqual(0, agent1.initializeAgentCalls);
             Assert.AreEqual(0, agent2.initializeAgentCalls);
             Assert.AreEqual(0, agent1.agentActionCalls);
 
             // agent1 was not enabled when the academy started
             // The agents have been initialized
-            Assert.AreEqual(0, agent1.agentResetCalls);
-            Assert.AreEqual(0, agent2.agentResetCalls);
+            Assert.AreEqual(0, agent1.agentOnEpisodeBeginCalls);
+            Assert.AreEqual(0, agent2.agentOnEpisodeBeginCalls);
             Assert.AreEqual(1, agent1.initializeAgentCalls);
             Assert.AreEqual(1, agent2.initializeAgentCalls);
             Assert.AreEqual(0, agent1.agentActionCalls);
 
             agent1.LazyInitialize();
 
-            var numberAgent1Reset = 0;
+            var numberAgent1Episodes = 0;
+            var numberAgent2Episodes = 0;
-                Assert.AreEqual(numberAgent1Reset, agent1.agentResetCalls);
-                // Agent2 is never reset since initialized after academy
-                Assert.AreEqual(0, agent2.agentResetCalls);
+                Assert.AreEqual(numberAgent1Episodes, agent1.agentOnEpisodeBeginCalls);
+                Assert.AreEqual(numberAgent2Episodes, agent2.agentOnEpisodeBeginCalls);
                 Assert.AreEqual(1, agent1.initializeAgentCalls);
                 Assert.AreEqual(numberAgent2Initialization, agent2.initializeAgentCalls);
                 Assert.AreEqual(i, agent1.agentActionCalls);
-                // Agent 1 resets at the first step
+                // Agent 1 starts a new episode at the first step
-                    numberAgent1Reset += 1;
+                    numberAgent1Episodes += 1;
+                    // Since Agent2 is initialized after the Academy has stepped, its OnEpisodeBegin should be called now.
+                    Assert.AreEqual(0, agent2.agentOnEpisodeBeginCalls);
+                    Assert.AreEqual(1, agent2.agentOnEpisodeBeginCalls);
+                    numberAgent2Episodes += 1;
                 }
 
                 // We are testing request decision and request actions when called
 
             agent2.LazyInitialize();
 
-            var numberAgent1Reset = 0;
-            var numberAgent2Reset = 0;
+            var numberAgent1Episodes = 0;
+            var numberAgent2Episodes = 0;
-            var agent2StepSinceReset = 0;
+            var agent2StepForEpisode = 0;
             for (var i = 0; i < 5000; i++)
             {
                 Assert.AreEqual(acaStepsSinceReset, aca.StepCount);
-
-                Assert.AreEqual(agent2StepSinceReset, agent2.StepCount);
-                Assert.AreEqual(numberAgent1Reset, agent1.agentResetCalls);
-                Assert.AreEqual(numberAgent2Reset, agent2.agentResetCalls);
+                Assert.AreEqual(numberAgent2Episodes, agent2.agentOnEpisodeBeginCalls);
+                Assert.AreEqual(agent2StepForEpisode, agent2.StepCount);
-                // Agent 2  and academy reset at the first step
+                // Agent 2 and academy reset at the first step
+                    Assert.AreEqual(numberAgent2Episodes, agent2.agentOnEpisodeBeginCalls);
-                    numberAgent2Reset += 1;
+                    numberAgent2Episodes += 1;
+                    Assert.AreEqual(numberAgent1Episodes, agent1.agentOnEpisodeBeginCalls);
+                    numberAgent1Episodes += 1;
+                    Assert.AreEqual(numberAgent1Episodes, agent1.agentOnEpisodeBeginCalls);
-                    agent1.Done();
-                    numberAgent1Reset += 1;
+                    Assert.AreEqual(numberAgent1Episodes, agent1.agentOnEpisodeBeginCalls);
+                    agent1.EndEpisode();
+                    numberAgent1Episodes += 1;
+                    Assert.AreEqual(numberAgent1Episodes, agent1.agentOnEpisodeBeginCalls);
-                // Resetting agent 2 regularly
+                // Ending the episode for agent 2 regularly
-                    agent2.Done();
-                    numberAgent2Reset += 1;
-                    agent2StepSinceReset = 0;
+                    Assert.AreEqual(numberAgent2Episodes, agent2.agentOnEpisodeBeginCalls);
+                    agent2.EndEpisode();
+                    numberAgent2Episodes += 1;
+                    agent2StepForEpisode = 0;
+                    Assert.AreEqual(numberAgent2Episodes, agent2.agentOnEpisodeBeginCalls);
                 }
                 // Request a decision for agent 2 regularly
                 if (i % 3 == 2)
                 }
 
                 acaStepsSinceReset += 1;
-                agent2StepSinceReset += 1;
+                agent2StepForEpisode += 1;
                 aca.EnvironmentStep();
             }
         }
             agent1.LazyInitialize();
             agent2.SetPolicy(new TestPolicy());
 
-            var expectedAgent1ActionSinceReset = 0;
+            var expectedAgent1ActionForEpisode = 0;
-                expectedAgent1ActionSinceReset += 1;
-                if (expectedAgent1ActionSinceReset == agent1.maxStep || i == 0)
+                expectedAgent1ActionForEpisode += 1;
+                if (expectedAgent1ActionForEpisode == agent1.maxStep || i == 0)
-                    expectedAgent1ActionSinceReset = 0;
+                    expectedAgent1ActionForEpisode = 0;
-                Assert.LessOrEqual(Mathf.Abs(expectedAgent1ActionSinceReset * 10.1f - agent1.GetCumulativeReward()), 0.05f);
+                Assert.LessOrEqual(Mathf.Abs(expectedAgent1ActionForEpisode * 10.1f - agent1.GetCumulativeReward()), 0.05f);
                 Assert.LessOrEqual(Mathf.Abs(i * 0.1f - agent2.GetCumulativeReward()), 0.05f);
 
                 agent1.AddReward(10f);
             agent1.LazyInitialize();
 
             var expectedAgentStepCount = 0;
-            var expectedResets = 0;
+            var expectedEpisodes = 0;
-            var expectedAgentActionSinceReset = 0;
+            var expectedAgentActionForEpisode = 0;
-            var expectedCollectObsCallsSinceReset = 0;
+            var expectedCollectObsCallsForEpisode = 0;
-                expectedAgentActionSinceReset += 1;
+                expectedAgentActionForEpisode += 1;
-                expectedCollectObsCallsSinceReset += 1;
+                expectedCollectObsCallsForEpisode += 1;
-                    expectedResets += 1;
+                    expectedEpisodes += 1;
-                    expectedAgentActionSinceReset = 0;
-                    expectedCollectObsCallsSinceReset = 0;
+                    expectedAgentActionForEpisode = 0;
+                    expectedCollectObsCallsForEpisode = 0;
-                Assert.AreEqual(expectedResets, agent1.agentResetCalls);
+                Assert.AreEqual(expectedEpisodes, agent1.agentOnEpisodeBeginCalls);
-                Assert.AreEqual(expectedAgentActionSinceReset, agent1.agentActionCallsSinceLastReset);
+                Assert.AreEqual(expectedAgentActionForEpisode, agent1.agentActionCallsForEpisode);
-                Assert.AreEqual(expectedCollectObsCallsSinceReset, agent1.collectObservationsCallsSinceLastReset);
+                Assert.AreEqual(expectedCollectObsCallsForEpisode, agent1.collectObservationsCallsForEpisode);
+        }
+
+        [Test]
+        public void TestHeuristicPolicyStepsSensors()
+        {
+            // Make sure that Agents with HeuristicPolicies step their sensors each Academy step.
+            var agentGo1 = new GameObject("TestAgent");
+            agentGo1.AddComponent<TestAgent>();
+            var agent1 = agentGo1.GetComponent<TestAgent>();
+            var aca = Academy.Instance;
+
+            var decisionRequester = agent1.gameObject.AddComponent<DecisionRequester>();
+            decisionRequester.DecisionPeriod = 1;
+            decisionRequester.Awake();
+
+            agent1.LazyInitialize();
+            Assert.AreEqual(agent1.GetPolicy().GetType(), typeof(HeuristicPolicy));
+
+            var numSteps = 10;
+            for (var i = 0; i < numSteps; i++)
+            {
+                aca.EnvironmentStep();
+            }
+            Assert.AreEqual(numSteps, agent1.heuristicCalls);
+            Assert.AreEqual(numSteps, agent1.sensor1.numWriteCalls);
+            Assert.AreEqual(numSteps, agent1.sensor2.numCompressedCalls);
+        }
+    }
+
+    [TestFixture]
+    public class TestOnEnableOverride
+    {
+        public class OnEnableAgent : Agent
+        {
+            public bool callBase;
+
+            protected override void OnEnable()
+            {
+                if (callBase)
+                    base.OnEnable();
+            }
+        }
+
+        static void _InnerAgentTestOnEnableOverride(bool callBase = false)
+        {
+            var go = new GameObject();
+            var agent = go.AddComponent<OnEnableAgent>();
+            agent.callBase = callBase;
+            var onEnable = typeof(OnEnableAgent).GetMethod("OnEnable", BindingFlags.NonPublic | BindingFlags.Instance);
+            var sendInfo = typeof(Agent).GetMethod("SendInfoToBrain", BindingFlags.NonPublic | BindingFlags.Instance);
+            Assert.NotNull(onEnable);
+            onEnable.Invoke(agent, null);
+            Assert.NotNull(sendInfo);
+            if (agent.callBase)
+            {
+                Assert.DoesNotThrow(() => sendInfo.Invoke(agent, null));
+            }
+            else
+            {
+                Assert.Throws<UnityAgentsException>(() =>
+                {
+                    try
+                    {
+                        sendInfo.Invoke(agent, null);
+                    }
+                    catch (TargetInvocationException e)
+                    {
+                        throw e.GetBaseException();
+                    }
+                });
+            }
+        }
+
+        [Test]
+        public void TestAgentCallBaseOnEnable()
+        {
+            _InnerAgentTestOnEnableOverride(true);
+        }
+
+        [Test]
+        public void TestAgentDontCallBaseOnEnable()
+        {
+            _InnerAgentTestOnEnableOverride();
         }
     }
 }

com.unity.ml-agents/Tests/Editor/ModelRunnerTest.cs

 using UnityEngine;
 using UnityEditor;
 using Barracuda;
-using MLAgents.InferenceBrain;
+using MLAgents.Inference;
+using MLAgents.Policies;
 
 namespace MLAgents.Tests
 {

com.unity.ml-agents/Tests/Editor/MultinomialTest.cs

 using NUnit.Framework;
-using MLAgents.InferenceBrain.Utils;
+using MLAgents.Inference.Utils;
 
 namespace MLAgents.Tests
 {

com.unity.ml-agents/Tests/Editor/ParameterLoaderTest.cs

 using UnityEngine;
 using UnityEditor;
 using Barracuda;
-using MLAgents.InferenceBrain;
+using MLAgents.Inference;
+using MLAgents.Policies;
 
 namespace MLAgents.Tests
 {

com.unity.ml-agents/Tests/Editor/RandomNormalTest.cs

 using System;
 using NUnit.Framework;
-using MLAgents.InferenceBrain.Utils;
+using MLAgents.Inference.Utils;
 
 namespace MLAgents.Tests
 {

com.unity.ml-agents/Tests/Editor/Sensor/WriterAdapterTests.cs

 using NUnit.Framework;
 using MLAgents.Sensors;
 using Barracuda;
-using MLAgents.InferenceBrain;
+using MLAgents.Inference;
 
 
 namespace MLAgents.Tests

com.unity.ml-agents/Tests/Editor/SideChannelTests.cs

 using NUnit.Framework;
 using System.Collections.Generic;
 using System.Text;
+using MLAgents.SideChannels;
 
 namespace MLAgents.Tests
 {
         {
             public List<int> messagesReceived = new List<int>();
 
-            public TestSideChannel() { 
-                ChannelId = new Guid("6afa2c06-4f82-11ea-b238-784f4387d1f7"); 
+            public TestSideChannel()
+            {
+                ChannelId = new Guid("6afa2c06-4f82-11ea-b238-784f4387d1f7");
-            public override void OnMessageReceived(byte[] data)
+            public override void OnMessageReceived(IncomingMessage msg)
-                messagesReceived.Add(BitConverter.ToInt32(data, 0));
+                messagesReceived.Add(msg.ReadInt32());
-            public void SendInt(int data)
+            public void SendInt(int value)
-                QueueMessageToSend(BitConverter.GetBytes(data));
+                using (var msg = new OutgoingMessage())
+                {
+                    msg.WriteInt32(value);
+                    QueueMessageToSend(msg);
+                }
             }
         }
 
             fakeData = RpcCommunicator.GetSideChannelMessage(dictSender);
             RpcCommunicator.ProcessSideChannelData(dictReceiver, fakeData);
             Assert.AreEqual(wasCalled, 1);
+
+            var keysA = propA.ListProperties();
+            Assert.AreEqual(2, keysA.Count);
+            Assert.IsTrue(keysA.Contains(k1));
+            Assert.IsTrue(keysA.Contains(k2));
+
+            var keysB = propA.ListProperties();
+            Assert.AreEqual(2, keysB.Count);
+            Assert.IsTrue(keysB.Contains(k1));
+            Assert.IsTrue(keysB.Contains(k2));
+        }
+
+        [Test]
+        public void TestOutgoingMessageRawBytes()
+        {
+            // Make sure that SetRawBytes resets the buffer correctly.
+            // Write 8 bytes (an int and float) then call SetRawBytes with 4 bytes
+            var msg = new OutgoingMessage();
+            msg.WriteInt32(42);
+            msg.WriteFloat32(1.0f);
+
+            var data = new byte[] { 1, 2, 3, 4 };
+            msg.SetRawBytes(data);
+
+            var result = msg.ToByteArray();
+            Assert.AreEqual(data, result);
+        }
+
+        [Test]
+        public void TestMessageReadWrites()
+        {
+            var boolVal = true;
+            var intVal = 1337;
+            var floatVal = 4.2f;
+            var floatListVal = new float[] { 1001, 1002 };
+            var stringVal = "mlagents!";
+
+            IncomingMessage incomingMsg;
+            using (var outgoingMsg = new OutgoingMessage())
+            {
+                outgoingMsg.WriteBoolean(boolVal);
+                outgoingMsg.WriteInt32(intVal);
+                outgoingMsg.WriteFloat32(floatVal);
+                outgoingMsg.WriteString(stringVal);
+                outgoingMsg.WriteFloatList(floatListVal);
+
+                incomingMsg = new IncomingMessage(outgoingMsg.ToByteArray());
+            }
+
+            Assert.AreEqual(boolVal, incomingMsg.ReadBoolean());
+            Assert.AreEqual(intVal, incomingMsg.ReadInt32());
+            Assert.AreEqual(floatVal, incomingMsg.ReadFloat32());
+            Assert.AreEqual(stringVal, incomingMsg.ReadString());
+            Assert.AreEqual(floatListVal, incomingMsg.ReadFloatList());
         }
     }
 }

com.unity.ml-agents/Tests/Editor/TensorUtilsTest.cs

 using System;
 using Barracuda;
-using MLAgents.InferenceBrain;
-using MLAgents.InferenceBrain.Utils;
+using MLAgents.Inference;
+using MLAgents.Inference.Utils;
 using NUnit.Framework;
 
 namespace MLAgents.Tests

com.unity.ml-agents/package.json

 {
 	"name": "com.unity.ml-agents",
 	"displayName":"ML Agents",
-	"version": "0.14.1-preview",
+	"version": "0.15.0-preview",
-		"com.unity.barracuda": "0.6.0-preview"
+		"com.unity.barracuda": "0.6.1-preview"
 	}
 }

config/sac_trainer_config.yaml

     summary_freq: 20000
 
 PushBlock:
-    max_steps: 1.5e7
+    max_steps: 2e6
     init_entcoef: 0.05
     hidden_units: 256
     summary_freq: 60000