-

config/ppo/3DBall.yaml

     trainer_type: ppo
     hyperparameters:
       batch_size: 64
-      buffer_size: 12000
+      buffer_size: 120
       learning_rate: 0.0003
       beta: 0.001
       epsilon: 0.2
         gamma: 0.99
         strength: 1.0
     keep_checkpoints: 5
-    max_steps: 500
+    max_steps: 5000
     time_horizon: 1000
     summary_freq: 12000
     threaded: true