PushJump modifications

ml-agents/mlagents/trainers/torch/layers.py

             if isinstance(layer, Swish):
                 activation = layer(activation)
             else:
-                activation = layer(activation) + self.goal_encoders[idx // 2](
+                activation = layer(activation) * self.goal_encoders[idx // 2](
                     goal_tensor
                 )
         return activation

Project/Assets/ML-Agents/Examples/PushJump/Scripts/WJPBAgent.cs

     public GameObject ground;
     public GameObject spawnArea;
     Bounds m_SpawnAreaBounds;
-
+    GoalSensorComponent goalSensor;
 
     public GameObject wallJumpGoal;
     public GameObject pushBlockGoal;
 
         sensor.AddObservation(agentPos / 20f);
         sensor.AddObservation(DoGroundCheck(true) ? 1 : 0);
-        sensor.AddObservation(m_GoalOneHot);
+        goalSensor = this.GetComponent<GoalSensorComponent>();
+        goalSensor.AddGoal(m_GoalOneHot);
     }
 
     /// <summary>
 
     public void MoveAgent(ActionSegment<int> act)
     {
-        AddReward(-0.0005f);
+        AddReward(-0.0002f);
         var smallGrounded = DoGroundCheck(true);
         var largeGrounded = DoGroundCheck(false);
 
                 SetReward(1f);
                 EndEpisode();
             }
+            else
+            {
+                SetReward(-1f);
+                EndEpisode();
+            }
         }
     }
 
         {
             SetReward(1f);
+            EndEpisode();
+        }
+        else
+        {
+            SetReward(-1f);
             EndEpisode();
         }
     }
             m_PushRenderer.material = goalMaterial;
         }
 
-        var height = m_ResetParams.GetWithDefault("big_wall_height", 8);
+        var height = m_ResetParams.GetWithDefault("big_wall_height", 4);
         var localScale = wall.transform.localScale;
         localScale = new Vector3(
             localScale.x,

config/ppo/PushJump.yaml

+behaviors:
+  WJPB:
+    trainer_type: ppo
+    hyperparameters:
+      batch_size: 128
+      buffer_size: 5000
+      learning_rate: 0.0003
+      beta: 0.01
+      epsilon: 0.2
+      lambd: 0.95
+      num_epoch: 3
+      learning_rate_schedule: linear
+    network_settings:
+      normalize: false
+      hidden_units: 64
+      num_layers: 2
+      vis_encode_type: simple
+    reward_signals:
+      extrinsic:
+        gamma: 0.99
+        strength: 1.0
+    keep_checkpoints: 5
+    max_steps: 2000000
+    time_horizon: 64
+    summary_freq: 30000
+    threaded: true

Project/Assets/ML-Agents/Examples/PushJump/TFModels/WJPB-32499994.onnx.meta

+fileFormatVersion: 2
+guid: 960b1933b6ec84fcb83c0c79668a4ee8
+ScriptedImporter:
+  fileIDToRecycleName:
+    11400000: main obj
+    11400002: model data
+  externalObjects: {}
+  userData: 
+  assetBundleName: 
+  assetBundleVariant: 
+  script: {fileID: 11500000, guid: 683b6cb6d0a474744822c888b46772c9, type: 3}
+  optimizeModel: 1
+  forceArbitraryBatchSize: 1
+  treatErrorsAsWarnings: 0

Project/Assets/ML-Agents/Examples/PushJump/TFModels/WJPB-4999918.onnx.meta

+fileFormatVersion: 2
+guid: 84bee0873d0be4a56b46a9884553728d
+ScriptedImporter:
+  fileIDToRecycleName:
+    11400000: main obj
+    11400002: model data
+  externalObjects: {}
+  userData: 
+  assetBundleName: 
+  assetBundleVariant: 
+  script: {fileID: 11500000, guid: 683b6cb6d0a474744822c888b46772c9, type: 3}
+  optimizeModel: 1
+  forceArbitraryBatchSize: 1
+  treatErrorsAsWarnings: 0

Project/Assets/ML-Agents/Examples/PushJump/TFModels/cloud.onnx.meta

+fileFormatVersion: 2
+guid: 15e1a204928a14cfa910f8c51edaf85f
+ScriptedImporter:
+  fileIDToRecycleName:
+    11400000: main obj
+    11400002: model data
+  externalObjects: {}
+  userData: 
+  assetBundleName: 
+  assetBundleVariant: 
+  script: {fileID: 11500000, guid: 683b6cb6d0a474744822c888b46772c9, type: 3}
+  optimizeModel: 1
+  forceArbitraryBatchSize: 1
+  treatErrorsAsWarnings: 0