Add exploration noise to rl training collector (#1481)

* Update vessel.py

Add exploration_noise=True  to training collector

* Update vessel.py

Reformat

qlib/rl/trainer/vessel.py

@@ -168,7 +168,9 @@ class TrainingVessel(TrainingVesselBase):
         self.policy.train()
 
         with vector_env.collector_guard():
-            collector = Collector(self.policy, vector_env, VectorReplayBuffer(self.buffer_size, len(vector_env)))
+            collector = Collector(
+                self.policy, vector_env, VectorReplayBuffer(self.buffer_size, len(vector_env)), exploration_noise=True
+            )
 
             # Number of episodes collected in each training iteration can be overridden by fast dev run.
             if self.trainer.fast_dev_run is not None: