8x batch size for cloud test

config/ppo/WalkerDynamic.yaml

 behaviors:
   WalkerDynamic:
     trainer: ppo
-    batch_size: 2048
+    batch_size: 16384 #2048
     beta: 0.005
     buffer_size: 163840 #20480
     epsilon: 0.2

config/trainer_config.yaml

             strength: 1.0
             gamma: 0.995
 
-# WalkerDynamic:
-#     normalize: true
-#     # learning_rate_schedule: constant
-#     num_epoch: 3
-#     time_horizon: 80
-#     batch_size: 512
-#     buffer_size: 5120
-#     max_steps: 2e7
-#     # max_steps: 1e8
-#     summary_freq: 30000
-#     # learning_rate: 1.0e-3
-#     num_layers: 3
-#     hidden_units: 512
-#     reward_signals:
-#         extrinsic:
-#             strength: 1.0
-#             gamma: 0.995
-
+    # learning_rate_schedule: constant
+    beta: 9.0e-3
+    # max_steps: 1e8
+    # learning_rate: 1.0e-3
     num_layers: 3
     hidden_units: 512
     reward_signals:
+
+# WalkerDynamic:
+#     normalize: true
+#     num_epoch: 3
+#     time_horizon: 1000
+#     batch_size: 2048
+#     buffer_size: 20480
+#     max_steps: 2e7
+#     summary_freq: 30000
+#     num_layers: 3
+#     hidden_units: 512
+#     reward_signals:
+#         extrinsic:
+#             strength: 1.0
+#             gamma: 0.995
 
 WalkerStatic:
     normalize: true