fix bug where epsilon_greedy is not passed to ppg_network_forward (#1447)

(cherry picked from commit 11318a98b839b37575052b2db233615b87cc91db)

Co-authored-by: Le Horizon <le.zhao+git@horizon.ai>

alf/algorithms/ppg_algorithm.py

@@ -199,8 +199,11 @@ class PPGAlgorithm(OffPolicyAlgorithm):
         return self._loss(info)
 
     def predict_step(self, inputs: TimeStep, state):
-        return ppg_network_forward(self._network, inputs, state,
-                                   self._predict_step_epsilon_greedy)
+        return ppg_network_forward(
+            self._network,
+            inputs,
+            state,
+            epsilon_greedy=self._predict_step_epsilon_greedy)
 
     def after_train_iter(self, experience, info: PPGTrainInfo):
         """Run auxiliary update if conditions are met