fix DiscretePolicyGradient set_state bug; upgrade pymaro version

maro/__misc__.py

@@ -2,6 +2,6 @@
 # Licensed under the MIT license.
 
 
-__version__ = "0.3.2a2"
+__version__ = "0.3.2a3"
 
 __data_version__ = "0.2"

maro/rl/policy/discrete_rl_policy.py

@@ -336,7 +336,7 @@ class DiscretePolicyGradient(DiscreteRLPolicy):
         }
 
     def set_state(self, policy_state: dict) -> None:
-        self._policy_net.set_state(policy_state)
+        self._policy_net.set_state(policy_state["net"])
         self._warmup = policy_state["policy"]["warmup"]
         self._call_count = policy_state["policy"]["call_count"]