diff --git a/ci/jenkins_tests/run_rllib_tests.sh b/ci/jenkins_tests/run_rllib_tests.sh
index 593a4f6c2..c65b045ba 100755
--- a/ci/jenkins_tests/run_rllib_tests.sh
+++ b/ci/jenkins_tests/run_rllib_tests.sh
@@ -60,6 +60,14 @@ docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     --env CartPole-v1 \
     --run PPO \
     --stop '{"training_iteration": 1}' \
+    --config '{"num_gpus": 0.1}' \
+    --ray-num-gpus 1
+
+docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    --env CartPole-v1 \
+    --run PPO \
+    --stop '{"training_iteration": 1}' \
     --config '{"kl_coeff": 1.0, "num_sgd_iter": 10, "lr": 1e-4, "sgd_minibatch_size": 64, "train_batch_size": 2000, "num_workers": 1, "use_gae": false, "batch_mode": "complete_episodes"}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
diff --git a/python/ray/tune/ray_trial_executor.py b/python/ray/tune/ray_trial_executor.py
index 52b8c63f5..6e031bd15 100644
--- a/python/ray/tune/ray_trial_executor.py
+++ b/python/ray/tune/ray_trial_executor.py
@@ -118,7 +118,7 @@ class RayTrialExecutor(TrialExecutor):
         # Clear the Trial's location (to be updated later on result)
         # since we don't know where the remote runner is placed.
         trial.set_location(Location())
-        logger.debug("Trial %s: Setting up new remote runner.", trial)
+        logger.info("Trial %s: Setting up new remote runner.", trial)
         # Logging for trials is handled centrally by TrialRunner, so
         # configure the remote runner to use a noop-logger.
         return cls.remote(config=trial.config, logger_creator=logger_creator)
diff --git a/rllib/evaluation/rollout_worker.py b/rllib/evaluation/rollout_worker.py
index 10beb3c50..ac9edd344 100644
--- a/rllib/evaluation/rollout_worker.py
+++ b/rllib/evaluation/rollout_worker.py
@@ -327,20 +327,14 @@ class RolloutWorker(EvaluatorInterface):
                 logger.info("Could not seed torch")
         if _has_tensorflow_graph(policy_dict) and not (tf and
                                                        tf.executing_eagerly()):
+            if (ray.is_initialized()
+                    and ray.worker._mode() != ray.worker.LOCAL_MODE
+                    and not ray.get_gpu_ids()):
+                logger.debug("Creating policy evaluation worker {}".format(
+                    worker_index) +
+                             " on CPU (please ignore any CUDA init errors)")
             if not tf:
                 raise ImportError("Could not import tensorflow")
-            if (ray.is_initialized()
-                    and ray.worker._mode() != ray.worker.LOCAL_MODE):
-                if not ray.get_gpu_ids():
-                    logger.debug(
-                        "Creating policy evaluation worker {}".format(
-                            worker_index) +
-                        " on CPU (please ignore any CUDA init errors)")
-                elif not tf.test.is_gpu_available():
-                    raise RuntimeError(
-                        "GPUs were assigned to this worker by Ray, but "
-                        "TensorFlow reports GPU acceleration is disabled. "
-                        "This could be due to a bad CUDA or TF installation.")
             with tf.Graph().as_default():
                 if tf_session_creator:
                     self.tf_sess = tf_session_creator()