[rllib] Fix DDPG example (#4973)

python/ray/rllib/policy/tf_policy.py

@@ -205,7 +205,7 @@ class TFPolicy(Policy):
                                                  self._grads_and_vars)
 
         if log_once("loss_used"):
-            logger.info(
+            logger.debug(
                 "These tensors were used in the loss_fn:\n\n{}\n".format(
                     summarize(self._loss_input_dict)))
 

python/ray/rllib/tuned_examples/halfcheetah-ddpg.yaml

@@ -47,7 +47,6 @@ halfcheetah-ddpg:
         # === Parallelism ===
         num_workers: 0
         num_gpus_per_worker: 0
-        optimizer_class: "SyncReplayOptimizer"
         per_worker_exploration: False
         worker_side_prioritization: False
 

python/ray/rllib/tuned_examples/mountaincarcontinuous-ddpg.yaml

@@ -47,7 +47,6 @@ mountaincarcontinuous-ddpg:
         # === Parallelism ===
         num_workers: 0
         num_gpus_per_worker: 0
-        optimizer_class: "SyncReplayOptimizer"
         per_worker_exploration: False
         worker_side_prioritization: False
 

python/ray/rllib/tuned_examples/pendulum-ddpg.yaml

@@ -47,7 +47,6 @@ pendulum-ddpg:
         # === Parallelism ===
         num_workers: 0
         num_gpus_per_worker: 0
-        optimizer_class: "SyncReplayOptimizer"
         per_worker_exploration: False
         worker_side_prioritization: False