[RLLib] DDPG (#1685)

python/ray/rllib/pg/policy.py

@@ -59,7 +59,7 @@ class PGPolicy():
     def compute_gradients(self, samples):
         info = {}
         feed_dict = {
-            self.x: samples["observations"],
+            self.x: samples["obs"],
             self.ac: samples["actions"],
             self.adv: samples["advantages"],
         }