diff --git a/python/ray/actor.py b/python/ray/actor.py
index 0d22ed9f5..49b928f58 100644
--- a/python/ray/actor.py
+++ b/python/ray/actor.py
@@ -27,7 +27,7 @@ def method(*args, **kwargs):
     .. code-block:: python
 
         @ray.remote
-        class Foo(object):
+        class Foo:
             @ray.method(num_return_vals=2)
             def bar(self):
                 return 1, 2
@@ -54,7 +54,7 @@ def method(*args, **kwargs):
 
 # Create objects to wrap method invocations. This is done so that we can
 # invoke methods with actor.method.remote() instead of actor.method().
-class ActorMethod(object):
+class ActorMethod:
     """A class used to invoke an actor method.
 
     Note: This class only keeps a weak ref to the actor, unless it has been
@@ -143,7 +143,7 @@ class ActorMethod(object):
             hardref=True)
 
 
-class ActorClassMetadata(object):
+class ActorClassMetadata:
     """Metadata for an actor class.
 
     Attributes:
@@ -234,7 +234,7 @@ class ActorClassMetadata(object):
                     method.__ray_invocation_decorator__)
 
 
-class ActorClass(object):
+class ActorClass:
     """An actor class.
 
     This is a decorated class. It can be used to create actors.
@@ -339,7 +339,7 @@ class ActorClass(object):
 
         actor_cls = self
 
-        class ActorOptionWrapper(object):
+        class ActorOptionWrapper:
             def remote(self, *args, **kwargs):
                 return actor_cls._remote(args=args, kwargs=kwargs, **options)
 
@@ -518,7 +518,7 @@ class ActorClass(object):
         return actor_handle
 
 
-class ActorHandle(object):
+class ActorHandle:
     """A handle to an actor.
 
     The fields in this class are prefixed with _ray_ to hide them from the user
diff --git a/python/ray/autoscaler/autoscaler.py b/python/ray/autoscaler/autoscaler.py
index e6b9e5c68..dd399b721 100644
--- a/python/ray/autoscaler/autoscaler.py
+++ b/python/ray/autoscaler/autoscaler.py
@@ -156,7 +156,7 @@ CLUSTER_CONFIG_SCHEMA = {
 }
 
 
-class LoadMetrics(object):
+class LoadMetrics:
     """Container for cluster load metrics.
 
     Metrics here are updated from raylet heartbeats. The autoscaler
@@ -355,7 +355,7 @@ class NodeLauncher(threading.Thread):
         logger.info(prefix + " {}".format(statement))
 
 
-class ConcurrentCounter(object):
+class ConcurrentCounter:
     def __init__(self):
         self._value = 0
         self._lock = threading.Lock()
@@ -377,7 +377,7 @@ class ConcurrentCounter(object):
             return self._value
 
 
-class StandardAutoscaler(object):
+class StandardAutoscaler:
     """The autoscaling control loop for a Ray cluster.
 
     There are two ways to start an autoscaling cluster: manually by running
diff --git a/python/ray/autoscaler/local/node_provider.py b/python/ray/autoscaler/local/node_provider.py
index 5996c0ca9..e4f9fc8bc 100644
--- a/python/ray/autoscaler/local/node_provider.py
+++ b/python/ray/autoscaler/local/node_provider.py
@@ -19,7 +19,7 @@ filelock_logger = logging.getLogger("filelock")
 filelock_logger.setLevel(logging.WARNING)
 
 
-class ClusterState(object):
+class ClusterState:
     def __init__(self, lock_path, save_path, provider_config):
         self.lock = RLock()
         self.file_lock = FileLock(lock_path)
diff --git a/python/ray/autoscaler/log_timer.py b/python/ray/autoscaler/log_timer.py
index 7506257d5..723d7a71c 100644
--- a/python/ray/autoscaler/log_timer.py
+++ b/python/ray/autoscaler/log_timer.py
@@ -8,7 +8,7 @@ import logging
 logger = logging.getLogger(__name__)
 
 
-class LogTimer(object):
+class LogTimer:
     def __init__(self, message):
         self._message = message
 
diff --git a/python/ray/autoscaler/node_provider.py b/python/ray/autoscaler/node_provider.py
index c9b0c9164..d8953e530 100644
--- a/python/ray/autoscaler/node_provider.py
+++ b/python/ray/autoscaler/node_provider.py
@@ -129,7 +129,7 @@ def get_default_config(provider_config):
     return defaults
 
 
-class NodeProvider(object):
+class NodeProvider:
     """Interface for getting and returning nodes from a Cloud.
 
     NodeProviders are namespaced by the `cluster_name` parameter; they only
diff --git a/python/ray/autoscaler/updater.py b/python/ray/autoscaler/updater.py
index ee93f9734..f200fe1f8 100644
--- a/python/ray/autoscaler/updater.py
+++ b/python/ray/autoscaler/updater.py
@@ -37,7 +37,7 @@ def with_interactive(cmd):
     return ["bash", "--login", "-c", "-i", quote(force_interactive + cmd)]
 
 
-class KubernetesCommandRunner(object):
+class KubernetesCommandRunner:
     def __init__(self, log_prefix, namespace, node_id, auth_config,
                  process_runner):
 
@@ -149,7 +149,7 @@ class KubernetesCommandRunner(object):
                                             self.node_id)
 
 
-class SSHCommandRunner(object):
+class SSHCommandRunner:
     def __init__(self, log_prefix, node_id, provider, auth_config,
                  cluster_name, process_runner, use_internal_ip):
 
@@ -284,7 +284,7 @@ class SSHCommandRunner(object):
                                           self.ssh_ip)
 
 
-class NodeUpdater(object):
+class NodeUpdater:
     """A process for syncing files and running init commands on a node."""
 
     def __init__(self,
diff --git a/python/ray/cluster_utils.py b/python/ray/cluster_utils.py
index 2a7cc0580..0c7988f80 100644
--- a/python/ray/cluster_utils.py
+++ b/python/ray/cluster_utils.py
@@ -13,7 +13,7 @@ from ray import ray_constants
 logger = logging.getLogger(__name__)
 
 
-class Cluster(object):
+class Cluster:
     def __init__(self,
                  initialize_head=False,
                  connect=False,
diff --git a/python/ray/dashboard/dashboard.py b/python/ray/dashboard/dashboard.py
index c7deba529..c43ac3400 100644
--- a/python/ray/dashboard/dashboard.py
+++ b/python/ray/dashboard/dashboard.py
@@ -72,7 +72,7 @@ def format_reply(reply):
             format_reply(item)
 
 
-class Dashboard(object):
+class Dashboard:
     """A dashboard process for monitoring Ray nodes.
 
     This dashboard is made up of a REST API which collates data published by
diff --git a/python/ray/experimental/actor_pool.py b/python/ray/experimental/actor_pool.py
index 34a97ba0b..4deef58f5 100644
--- a/python/ray/experimental/actor_pool.py
+++ b/python/ray/experimental/actor_pool.py
@@ -5,7 +5,7 @@ from __future__ import print_function
 import ray
 
 
-class ActorPool(object):
+class ActorPool:
     """Utility class to operate on a fixed pool of actors.
 
     Arguments:
diff --git a/python/ray/experimental/array/distributed/core.py b/python/ray/experimental/array/distributed/core.py
index 992d4375c..af758d528 100644
--- a/python/ray/experimental/array/distributed/core.py
+++ b/python/ray/experimental/array/distributed/core.py
@@ -9,7 +9,7 @@ import ray
 BLOCK_SIZE = 10
 
 
-class DistArray(object):
+class DistArray:
     def __init__(self, shape, objectids=None):
         self.shape = shape
         self.ndim = len(shape)
diff --git a/python/ray/experimental/async_api.py b/python/ray/experimental/async_api.py
index 2a3a6fad2..5cbbc929e 100644
--- a/python/ray/experimental/async_api.py
+++ b/python/ray/experimental/async_api.py
@@ -15,7 +15,7 @@ transport = None
 protocol = None
 
 
-class _ThreadSafeProxy(object):
+class _ThreadSafeProxy:
     """This class is used to create a thread-safe proxy for a given object.
         Every method call will be guarded with a lock.
 
diff --git a/python/ray/experimental/gcs_flush_policy.py b/python/ray/experimental/gcs_flush_policy.py
index 1d1717f03..36e5ae701 100644
--- a/python/ray/experimental/gcs_flush_policy.py
+++ b/python/ray/experimental/gcs_flush_policy.py
@@ -9,7 +9,7 @@ import ray
 import ray.cloudpickle as pickle
 
 
-class GcsFlushPolicy(object):
+class GcsFlushPolicy:
     """Experimental: a policy to control GCS flushing.
 
     Used by Monitor to enable automatic control of memory usage.
diff --git a/python/ray/experimental/multiprocessing/pool.py b/python/ray/experimental/multiprocessing/pool.py
index f9184c1b6..e0e6b06ec 100644
--- a/python/ray/experimental/multiprocessing/pool.py
+++ b/python/ray/experimental/multiprocessing/pool.py
@@ -105,7 +105,7 @@ class ResultThread(threading.Thread):
             raise TimeoutError
 
 
-class AsyncResult(object):
+class AsyncResult:
     """An asynchronous interface to task results.
 
     This should not be constructed directly.
@@ -168,7 +168,7 @@ class AsyncResult(object):
         return not self._result_thread.got_error()
 
 
-class IMapIterator(object):
+class IMapIterator:
     """Base class for OrderedIMapIterator and UnorderedIMapIterator."""
 
     def __init__(self, pool, func, iterable, chunksize=None):
@@ -273,7 +273,7 @@ class UnorderedIMapIterator(IMapIterator):
 
 
 @ray.remote
-class PoolActor(object):
+class PoolActor:
     """Actor used to process tasks submitted to a Pool."""
 
     def __init__(self, initializer=None, initargs=None):
@@ -298,7 +298,7 @@ class PoolActor(object):
 
 
 # https://docs.python.org/3/library/multiprocessing.html#module-multiprocessing.pool
-class Pool(object):
+class Pool:
     """A pool of actor processes that is used to process tasks in parallel.
 
     Args:
diff --git a/python/ray/experimental/no_return.py b/python/ray/experimental/no_return.py
index 8903942d9..e9dc0e754 100644
--- a/python/ray/experimental/no_return.py
+++ b/python/ray/experimental/no_return.py
@@ -3,7 +3,7 @@ from __future__ import division
 from __future__ import print_function
 
 
-class NoReturn(object):
+class NoReturn:
     """Do not store the return value in the object store.
 
     If a task returns this object, then Ray will not store this object in the
diff --git a/python/ray/experimental/queue.py b/python/ray/experimental/queue.py
index d2cc9f434..0102e36d9 100644
--- a/python/ray/experimental/queue.py
+++ b/python/ray/experimental/queue.py
@@ -16,7 +16,7 @@ class Full(Exception):
     pass
 
 
-class Queue(object):
+class Queue:
     """Queue implementation on Ray.
 
     Args:
@@ -132,7 +132,7 @@ class Queue(object):
 
 
 @ray.remote
-class _QueueActor(object):
+class _QueueActor:
     def __init__(self, maxsize):
         self.maxsize = maxsize
         self._init(maxsize)
diff --git a/python/ray/experimental/serve/tests/test_metric.py b/python/ray/experimental/serve/tests/test_metric.py
index fde360d00..54e0ac6db 100644
--- a/python/ray/experimental/serve/tests/test_metric.py
+++ b/python/ray/experimental/serve/tests/test_metric.py
@@ -8,7 +8,7 @@ from ray.experimental.serve.metric import MetricMonitor
 @pytest.fixture(scope="session")
 def start_target_actor(ray_instance):
     @ray.remote
-    class Target(object):
+    class Target:
         def __init__(self):
             self.counter_value = 0
 
diff --git a/python/ray/experimental/sgd/pytorch/pytorch_runner.py b/python/ray/experimental/sgd/pytorch/pytorch_runner.py
index bee87d516..4f3db0e22 100644
--- a/python/ray/experimental/sgd/pytorch/pytorch_runner.py
+++ b/python/ray/experimental/sgd/pytorch/pytorch_runner.py
@@ -17,7 +17,7 @@ from ray.experimental.sgd import utils
 logger = logging.getLogger(__name__)
 
 
-class PyTorchRunner(object):
+class PyTorchRunner:
     """Manages a PyTorch model for training."""
 
     def __init__(self,
diff --git a/python/ray/experimental/sgd/pytorch/pytorch_trainer.py b/python/ray/experimental/sgd/pytorch/pytorch_trainer.py
index c73636f81..1eda6e56f 100644
--- a/python/ray/experimental/sgd/pytorch/pytorch_trainer.py
+++ b/python/ray/experimental/sgd/pytorch/pytorch_trainer.py
@@ -21,7 +21,7 @@ from ray.experimental.sgd.pytorch.pytorch_runner import PyTorchRunner
 logger = logging.getLogger(__name__)
 
 
-class PyTorchTrainer(object):
+class PyTorchTrainer:
     """Train a PyTorch model using distributed PyTorch.
 
     Launches a set of actors which connect via distributed PyTorch and
diff --git a/python/ray/experimental/sgd/pytorch/utils.py b/python/ray/experimental/sgd/pytorch/utils.py
index 22c98d0bf..4e60df2a4 100644
--- a/python/ray/experimental/sgd/pytorch/utils.py
+++ b/python/ray/experimental/sgd/pytorch/utils.py
@@ -108,7 +108,7 @@ def validate(model, val_iterator, criterion, config):
     return stats
 
 
-class AverageMeter(object):
+class AverageMeter:
     """Computes and stores the average and current value."""
 
     def __init__(self):
diff --git a/python/ray/experimental/sgd/tf/tf_runner.py b/python/ray/experimental/sgd/tf/tf_runner.py
index 3e9d2b9b9..13903c073 100644
--- a/python/ray/experimental/sgd/tf/tf_runner.py
+++ b/python/ray/experimental/sgd/tf/tf_runner.py
@@ -20,7 +20,7 @@ def _try_import_strategy():
     return tf.distribute.experimental.MultiWorkerMirroredStrategy
 
 
-class TFRunner(object):
+class TFRunner:
     """Manages a TensorFlow model for training."""
 
     def __init__(self, model_creator, data_creator, config=None,
diff --git a/python/ray/experimental/sgd/tf/tf_trainer.py b/python/ray/experimental/sgd/tf/tf_trainer.py
index 28cb831b7..eed8f7f72 100644
--- a/python/ray/experimental/sgd/tf/tf_trainer.py
+++ b/python/ray/experimental/sgd/tf/tf_trainer.py
@@ -16,7 +16,7 @@ from ray.experimental.sgd.tf.tf_runner import TFRunner
 logger = logging.getLogger(__name__)
 
 
-class TFTrainer(object):
+class TFTrainer:
     def __init__(self,
                  model_creator,
                  data_creator,
diff --git a/python/ray/experimental/sgd/utils.py b/python/ray/experimental/sgd/utils.py
index 005f80e4c..ce6ab7c6f 100644
--- a/python/ray/experimental/sgd/utils.py
+++ b/python/ray/experimental/sgd/utils.py
@@ -8,7 +8,7 @@ import socket
 import time
 
 
-class TimerStat(object):
+class TimerStat:
     """A running stat for conveniently logging the duration of a code block.
 
     Note that this class is *not* thread-safe.
@@ -108,7 +108,7 @@ def find_free_port():
         return s.getsockname()[1]
 
 
-class AverageMeter(object):
+class AverageMeter:
     """Computes and stores the average and current value."""
 
     def __init__(self):
diff --git a/python/ray/experimental/signal.py b/python/ray/experimental/signal.py
index 147af8a0a..d7d4ef56b 100644
--- a/python/ray/experimental/signal.py
+++ b/python/ray/experimental/signal.py
@@ -18,7 +18,7 @@ ACTOR_DIED_STR = "ACTOR_DIED_SIGNAL"
 logger = logging.getLogger(__name__)
 
 
-class Signal(object):
+class Signal:
     """Base class for Ray signals."""
     pass
 
diff --git a/python/ray/experimental/tf_utils.py b/python/ray/experimental/tf_utils.py
index 900cc948b..99c657562 100644
--- a/python/ray/experimental/tf_utils.py
+++ b/python/ray/experimental/tf_utils.py
@@ -22,7 +22,7 @@ def unflatten(vector, shapes):
     return arrays
 
 
-class TensorFlowVariables(object):
+class TensorFlowVariables:
     """A class used to set and get weights for Tensorflow networks.
 
     Attributes:
diff --git a/python/ray/function_manager.py b/python/ray/function_manager.py
index bd9b2bf8c..ece7ccf96 100644
--- a/python/ray/function_manager.py
+++ b/python/ray/function_manager.py
@@ -39,7 +39,7 @@ FunctionExecutionInfo = namedtuple("FunctionExecutionInfo",
 logger = logging.getLogger(__name__)
 
 
-class FunctionDescriptor(object):
+class FunctionDescriptor:
     """A class used to describe a python function.
 
     Attributes:
@@ -259,7 +259,7 @@ class FunctionDescriptor(object):
         return len(self._class_name) > 0
 
 
-class FunctionActorManager(object):
+class FunctionActorManager:
     """A class used to export/load remote functions and actors.
 
     Attributes:
@@ -658,7 +658,7 @@ class FunctionActorManager(object):
                     class_name))
 
     def _create_fake_actor_class(self, actor_class_name, actor_method_names):
-        class TemporaryActor(object):
+        class TemporaryActor:
             pass
 
         def temporary_actor_method(*args, **kwargs):
diff --git a/python/ray/import_thread.py b/python/ray/import_thread.py
index 0e564bc66..364982cf2 100644
--- a/python/ray/import_thread.py
+++ b/python/ray/import_thread.py
@@ -19,7 +19,7 @@ import logging
 logger = logging.getLogger(__name__)
 
 
-class ImportThread(object):
+class ImportThread:
     """A thread used to import exports from the driver or other workers.
 
     Note: The driver also has an import thread, which is used only to import
diff --git a/python/ray/local_mode_manager.py b/python/ray/local_mode_manager.py
index 11d74784a..d86e75237 100644
--- a/python/ray/local_mode_manager.py
+++ b/python/ray/local_mode_manager.py
@@ -35,7 +35,7 @@ class LocalModeObjectID(ObjectID):
         return new
 
 
-class LocalModeManager(object):
+class LocalModeManager:
     """Used to emulate remote operations when running in local mode."""
 
     def __init__(self):
diff --git a/python/ray/log_monitor.py b/python/ray/log_monitor.py
index 37239950e..645200e93 100644
--- a/python/ray/log_monitor.py
+++ b/python/ray/log_monitor.py
@@ -22,7 +22,7 @@ import ray.utils
 logger = logging.getLogger(__name__)
 
 
-class LogFileInfo(object):
+class LogFileInfo:
     def __init__(self,
                  filename=None,
                  size_when_last_opened=None,
@@ -37,7 +37,7 @@ class LogFileInfo(object):
         self.worker_pid = None
 
 
-class LogMonitor(object):
+class LogMonitor:
     """A monitor process for monitoring Ray log files.
 
     This class mantains a list of open files and a list of closed log files. We
diff --git a/python/ray/memory_monitor.py b/python/ray/memory_monitor.py
index a6eabf667..3928ecb8e 100644
--- a/python/ray/memory_monitor.py
+++ b/python/ray/memory_monitor.py
@@ -65,7 +65,7 @@ class RayOutOfMemoryError(Exception):
                 "these memory limits to a lower value.")
 
 
-class MemoryMonitor(object):
+class MemoryMonitor:
     """Helper class for raising errors on low memory.
 
     This presents a much cleaner error message to users than what would happen
diff --git a/python/ray/monitor.py b/python/ray/monitor.py
index caa02a82c..6658a708a 100644
--- a/python/ray/monitor.py
+++ b/python/ray/monitor.py
@@ -23,7 +23,7 @@ from ray.utils import (binary_to_hex, binary_to_object_id, binary_to_task_id,
 logger = logging.getLogger(__name__)
 
 
-class Monitor(object):
+class Monitor:
     """A monitor for Ray processes.
 
     The monitor is in charge of cleaning up the tables in the global state
diff --git a/python/ray/node.py b/python/ray/node.py
index 811709048..4b7838730 100644
--- a/python/ray/node.py
+++ b/python/ray/node.py
@@ -30,7 +30,7 @@ logger = logging.getLogger(__name__)
 SESSION_LATEST = "session_latest"
 
 
-class Node(object):
+class Node:
     """An encapsulation of the Ray processes on a single node.
 
     This class is responsible for starting Ray processes and killing them,
@@ -882,7 +882,7 @@ class Node(object):
         return not any(self.dead_processes())
 
 
-class LocalNode(object):
+class LocalNode:
     """Imitate the node that manages the processes in local mode."""
 
     def kill_all_processes(self, *args, **kwargs):
diff --git a/python/ray/parameter.py b/python/ray/parameter.py
index 65d02805f..e88cacc60 100644
--- a/python/ray/parameter.py
+++ b/python/ray/parameter.py
@@ -10,7 +10,7 @@ from packaging import version
 import ray.ray_constants as ray_constants
 
 
-class RayParams(object):
+class RayParams:
     """A class used to store the parameters used by Ray.
 
     Attributes:
diff --git a/python/ray/profiling.py b/python/ray/profiling.py
index e6e36f953..b39410df5 100644
--- a/python/ray/profiling.py
+++ b/python/ray/profiling.py
@@ -5,7 +5,7 @@ from __future__ import print_function
 import ray
 
 
-class _NullLogSpan(object):
+class _NullLogSpan:
     """A log span context manager that does nothing"""
 
     def __enter__(self):
diff --git a/python/ray/projects/scripts.py b/python/ray/projects/scripts.py
index 028f1ac6f..bd36a79bb 100644
--- a/python/ray/projects/scripts.py
+++ b/python/ray/projects/scripts.py
@@ -147,7 +147,7 @@ def load_project_or_throw():
             "`ray project validate` to inspect the error.")
 
 
-class SessionRunner(object):
+class SessionRunner:
     """Class for setting up a session and executing commands in it."""
 
     def __init__(self, session_name=None):
diff --git a/python/ray/ray_perf.py b/python/ray/ray_perf.py
index 99ca4d886..bcda18792 100644
--- a/python/ray/ray_perf.py
+++ b/python/ray/ray_perf.py
@@ -11,7 +11,7 @@ filter_pattern = os.environ.get("TESTS_TO_RUN", "")
 
 
 @ray.remote(num_cpus=0)
-class Actor(object):
+class Actor:
     def small_value(self):
         return b"ok"
 
@@ -23,7 +23,7 @@ class Actor(object):
 
 
 @ray.remote(num_cpus=0)
-class Client(object):
+class Client:
     def __init__(self, servers):
         if not isinstance(servers, list):
             servers = [servers]
diff --git a/python/ray/remote_function.py b/python/ray/remote_function.py
index bb869d4ee..c6ff8b276 100644
--- a/python/ray/remote_function.py
+++ b/python/ray/remote_function.py
@@ -21,7 +21,7 @@ DEFAULT_REMOTE_FUNCTION_NUM_TASK_RETRIES = 3
 logger = logging.getLogger(__name__)
 
 
-class RemoteFunction(object):
+class RemoteFunction:
     """A remote function.
 
     This is a decorated function. It can be used to spawn tasks.
@@ -133,7 +133,7 @@ class RemoteFunction(object):
 
         func_cls = self
 
-        class FuncWrapper(object):
+        class FuncWrapper:
             def remote(self, *args, **kwargs):
                 return func_cls._remote(args=args, kwargs=kwargs, **options)
 
diff --git a/python/ray/reporter.py b/python/ray/reporter.py
index be009172a..65c44d617 100644
--- a/python/ray/reporter.py
+++ b/python/ray/reporter.py
@@ -56,7 +56,7 @@ def to_posix_time(dt):
     return (dt - datetime.datetime(1970, 1, 1)).total_seconds()
 
 
-class Reporter(object):
+class Reporter:
     """A monitor process for monitoring Ray nodes.
 
     Attributes:
diff --git a/python/ray/runtime_context.py b/python/ray/runtime_context.py
index 0c1c88e00..dec103ca9 100644
--- a/python/ray/runtime_context.py
+++ b/python/ray/runtime_context.py
@@ -5,7 +5,7 @@ from __future__ import print_function
 import ray.worker
 
 
-class RuntimeContext(object):
+class RuntimeContext:
     """A class used for getting runtime context."""
 
     def __init__(self, worker=None):
diff --git a/python/ray/serialization.py b/python/ray/serialization.py
index 7e0c07ea0..2fbe13683 100644
--- a/python/ray/serialization.py
+++ b/python/ray/serialization.py
@@ -40,7 +40,7 @@ class DeserializationError(Exception):
     pass
 
 
-class SerializedObject(object):
+class SerializedObject:
     def __init__(self, metadata):
         self._metadata = metadata
 
@@ -132,7 +132,7 @@ def _try_to_compute_deterministic_class_id(cls, depth=5):
     return hashlib.sha1(new_class_id).digest()
 
 
-class SerializationContext(object):
+class SerializationContext:
     """Initialize the serialization library.
 
     This defines a custom serializer for object IDs and also tells ray to
diff --git a/python/ray/state.py b/python/ray/state.py
index ad56392e6..8f6e50688 100644
--- a/python/ray/state.py
+++ b/python/ray/state.py
@@ -121,7 +121,7 @@ def _parse_resource_table(redis_client, client_id):
     return resources
 
 
-class GlobalState(object):
+class GlobalState:
     """A class used to interface with the Ray control state.
 
     # TODO(zongheng): In the future move this to use Ray's redis module in the
diff --git a/python/ray/tests/py3_test.py b/python/ray/tests/py3_test.py
index abc547b72..3a7cd864f 100644
--- a/python/ray/tests/py3_test.py
+++ b/python/ray/tests/py3_test.py
@@ -100,7 +100,7 @@ def test_args_intertwined(ray_start_regular):
 
 def test_asyncio_actor(ray_start_regular_shared):
     @ray.remote
-    class AsyncBatcher(object):
+    class AsyncBatcher:
         def __init__(self):
             self.batch = []
             self.event = asyncio.Event()
diff --git a/python/ray/tests/test_actor.py b/python/ray/tests/test_actor.py
index 77a9b900f..9c43c65fc 100644
--- a/python/ray/tests/test_actor.py
+++ b/python/ray/tests/test_actor.py
@@ -22,7 +22,7 @@ from ray.experimental.internal_kv import _internal_kv_get, _internal_kv_put
 
 def test_actor_init_error_propagated(ray_start_regular):
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self, error=False):
             if error:
                 raise Exception("oops")
@@ -40,7 +40,7 @@ def test_actor_init_error_propagated(ray_start_regular):
 
 def test_keyword_args(ray_start_regular):
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self, arg0, arg1=1, arg2="a"):
             self.arg0 = arg0
             self.arg1 = arg1
@@ -89,7 +89,7 @@ def test_keyword_args(ray_start_regular):
 
 def test_variable_number_of_args(ray_start_regular):
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self, arg0, arg1=1, *args):
             self.arg0 = arg0
             self.arg1 = arg1
@@ -113,7 +113,7 @@ def test_variable_number_of_args(ray_start_regular):
         2, 3, 1, 2, 3, 4)) == (3, 5, ("a", "b", "c", "d"), (1, 2, 3, 4))
 
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self, *args):
             self.args = args
 
@@ -130,7 +130,7 @@ def test_variable_number_of_args(ray_start_regular):
 
 def test_no_args(ray_start_regular):
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self):
             pass
 
@@ -144,7 +144,7 @@ def test_no_args(ray_start_regular):
 def test_no_constructor(ray_start_regular):
     # If no __init__ method is provided, that should not be a problem.
     @ray.remote
-    class Actor(object):
+    class Actor:
         def get_values(self):
             pass
 
@@ -153,12 +153,12 @@ def test_no_constructor(ray_start_regular):
 
 
 def test_custom_classes(ray_start_regular):
-    class Foo(object):
+    class Foo:
         def __init__(self, x):
             self.x = x
 
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self, f2):
             self.f1 = Foo(1)
             self.f2 = f2
@@ -180,13 +180,13 @@ def test_custom_classes(ray_start_regular):
 
 
 def test_actor_class_attributes(ray_start_regular):
-    class Grandparent(object):
+    class Grandparent:
         GRANDPARENT = 2
 
     class Parent1(Grandparent):
         PARENT1 = 6
 
-    class Parent2(object):
+    class Parent2:
         PARENT2 = 7
 
     @ray.remote
@@ -216,7 +216,7 @@ def test_caching_actors(shutdown_only):
     # Test defining actors before ray.init() has been called.
 
     @ray.remote
-    class Foo(object):
+    class Foo:
         def __init__(self):
             pass
 
@@ -240,7 +240,7 @@ def test_decorator_args(ray_start_regular):
     with pytest.raises(Exception):
 
         @ray.remote()
-        class Actor(object):
+        class Actor:
             def __init__(self):
                 pass
 
@@ -248,7 +248,7 @@ def test_decorator_args(ray_start_regular):
     with pytest.raises(Exception):
 
         @ray.remote(invalid_kwarg=0)  # noqa: F811
-        class Actor(object):
+        class Actor:
             def __init__(self):
                 pass
 
@@ -256,32 +256,32 @@ def test_decorator_args(ray_start_regular):
     with pytest.raises(Exception):
 
         @ray.remote(num_cpus=0, invalid_kwarg=0)  # noqa: F811
-        class Actor(object):
+        class Actor:
             def __init__(self):
                 pass
 
     # This is a valid way of using the decorator.
     @ray.remote(num_cpus=1)  # noqa: F811
-    class Actor(object):
+    class Actor:
         def __init__(self):
             pass
 
     # This is a valid way of using the decorator.
     @ray.remote(num_gpus=1)  # noqa: F811
-    class Actor(object):
+    class Actor:
         def __init__(self):
             pass
 
     # This is a valid way of using the decorator.
     @ray.remote(num_cpus=1, num_gpus=1)  # noqa: F811
-    class Actor(object):
+    class Actor:
         def __init__(self):
             pass
 
 
 def test_random_id_generation(ray_start_regular):
     @ray.remote
-    class Foo(object):
+    class Foo:
         def __init__(self):
             pass
 
@@ -299,7 +299,7 @@ def test_random_id_generation(ray_start_regular):
 
 def test_actor_class_name(ray_start_regular):
     @ray.remote
-    class Foo(object):
+    class Foo:
         def __init__(self):
             pass
 
@@ -314,7 +314,7 @@ def test_actor_class_name(ray_start_regular):
 
 
 def test_actor_inheritance(ray_start_regular):
-    class NonActorBase(object):
+    class NonActorBase:
         def __init__(self):
             pass
 
@@ -342,7 +342,7 @@ def test_actor_inheritance(ray_start_regular):
 
 def test_multiple_return_values(ray_start_regular):
     @ray.remote
-    class Foo(object):
+    class Foo:
         def method0(self):
             return 1
 
@@ -375,7 +375,7 @@ def test_multiple_return_values(ray_start_regular):
 
 def test_define_actor(ray_start_regular):
     @ray.remote
-    class Test(object):
+    class Test:
         def __init__(self, x):
             self.x = x
 
@@ -395,7 +395,7 @@ def test_actor_deletion(ray_start_regular):
     # destructor is called.
 
     @ray.remote
-    class Actor(object):
+    class Actor:
         def getpid(self):
             return os.getpid()
 
@@ -413,7 +413,7 @@ def test_actor_deletion(ray_start_regular):
 
 def test_actor_method_deletion(ray_start_regular):
     @ray.remote
-    class Actor(object):
+    class Actor:
         def method(self):
             return 1
 
@@ -426,7 +426,7 @@ def test_actor_method_deletion(ray_start_regular):
 
 def test_multiple_actors(ray_start_regular):
     @ray.remote
-    class Counter(object):
+    class Counter:
         def __init__(self, value):
             self.value = value
 
@@ -479,7 +479,7 @@ def test_remote_function_within_actor(ray_start_10_cpus):
         return ray.get(f.remote(x))
 
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self, x):
             self.x = x
             self.y = val2
@@ -515,13 +515,13 @@ def test_define_actor_within_actor(ray_start_10_cpus):
     # Make sure we can use remote funtions within actors.
 
     @ray.remote
-    class Actor1(object):
+    class Actor1:
         def __init__(self, x):
             self.x = x
 
         def new_actor(self, z):
             @ray.remote
-            class Actor2(object):
+            class Actor2:
                 def __init__(self, x):
                     self.x = x
 
@@ -542,7 +542,7 @@ def test_use_actor_within_actor(ray_start_10_cpus):
     # Make sure we can use actors within actors.
 
     @ray.remote
-    class Actor1(object):
+    class Actor1:
         def __init__(self, x):
             self.x = x
 
@@ -550,7 +550,7 @@ def test_use_actor_within_actor(ray_start_10_cpus):
             return self.x
 
     @ray.remote
-    class Actor2(object):
+    class Actor2:
         def __init__(self, x, y):
             self.x = x
             self.actor1 = Actor1.remote(y)
@@ -568,7 +568,7 @@ def test_define_actor_within_remote_function(ray_start_10_cpus):
     @ray.remote
     def f(x, n):
         @ray.remote
-        class Actor1(object):
+        class Actor1:
             def __init__(self, x):
                 self.x = x
 
@@ -587,7 +587,7 @@ def test_use_actor_within_remote_function(ray_start_10_cpus):
     # Make sure we can create and use actors within remote funtions.
 
     @ray.remote
-    class Actor1(object):
+    class Actor1:
         def __init__(self, x):
             self.x = x
 
@@ -618,7 +618,7 @@ def test_actor_import_counter(ray_start_10_cpus):
     @ray.remote
     def g():
         @ray.remote
-        class Actor(object):
+        class Actor:
             def __init__(self):
                 # This should use the last version of f.
                 self.x = ray.get(f.remote())
@@ -636,7 +636,7 @@ def test_inherit_actor_from_class(ray_start_regular):
     # Make sure we can define an actor by inheriting from a regular class.
     # Note that actors cannot inherit from other actors.
 
-    class Foo(object):
+    class Foo:
         def __init__(self, x):
             self.x = x
 
@@ -663,7 +663,7 @@ def test_remote_functions_not_scheduled_on_actors(ray_start_regular):
     # Make sure that regular remote functions are not scheduled on actors.
 
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self):
             pass
 
@@ -683,7 +683,7 @@ def test_remote_functions_not_scheduled_on_actors(ray_start_regular):
 
 def test_actors_on_nodes_with_no_cpus(ray_start_no_cpu):
     @ray.remote
-    class Foo(object):
+    class Foo:
         def method(self):
             pass
 
@@ -700,7 +700,7 @@ def test_actor_load_balancing(ray_start_cluster):
     ray.init(address=cluster.address)
 
     @ray.remote
-    class Actor1(object):
+    class Actor1:
         def __init__(self):
             pass
 
@@ -743,7 +743,7 @@ def test_actor_lifetime_load_balancing(ray_start_cluster):
     ray.init(address=cluster.address)
 
     @ray.remote(num_cpus=1)
-    class Actor(object):
+    class Actor:
         def __init__(self):
             pass
 
@@ -759,7 +759,7 @@ def test_exception_raised_when_actor_node_dies(ray_start_cluster_head):
     remote_node = cluster.add_node()
 
     @ray.remote(max_reconstructions=0)
-    class Counter(object):
+    class Counter:
         def __init__(self):
             self.x = 0
 
@@ -799,7 +799,7 @@ def test_actor_init_fails(ray_start_cluster_head):
     remote_node = cluster.add_node()
 
     @ray.remote(max_reconstructions=1)
-    class Counter(object):
+    class Counter:
         def __init__(self):
             self.x = 0
 
@@ -825,7 +825,7 @@ def test_reconstruction_suppression(ray_start_cluster_head):
     worker_nodes = [cluster.add_node() for _ in range(num_nodes)]
 
     @ray.remote(max_reconstructions=1)
-    class Counter(object):
+    class Counter:
         def __init__(self):
             self.x = 0
 
@@ -863,7 +863,7 @@ def setup_counter_actor(test_checkpoint=False,
         checkpoint_interval = 5
 
     @ray.remote(checkpoint_interval=checkpoint_interval)
-    class Counter(object):
+    class Counter:
         _resume_exception = resume_exception
 
         def __init__(self, save_exception):
@@ -1031,7 +1031,7 @@ def _test_nondeterministic_reconstruction(
         cluster, num_forks, num_items_per_fork, num_forks_to_wait):
     # Make a shared queue.
     @ray.remote
-    class Queue(object):
+    class Queue:
         def __init__(self):
             self.queue = []
 
@@ -1115,7 +1115,7 @@ def setup_queue_actor():
     ray.init(num_cpus=1, object_store_memory=int(150 * 1024 * 1024))
 
     @ray.remote
-    class Queue(object):
+    class Queue:
         def __init__(self):
             self.queue = []
 
@@ -1274,7 +1274,7 @@ def test_garbage_collection(setup_queue_actor):
 
 def test_calling_put_on_actor_handle(ray_start_regular):
     @ray.remote
-    class Counter(object):
+    class Counter:
         def __init__(self):
             self.x = 0
 
@@ -1308,7 +1308,7 @@ def test_calling_put_on_actor_handle(ray_start_regular):
 
 def test_pickling_actor_handle(ray_start_regular):
     @ray.remote
-    class Foo(object):
+    class Foo:
         def method(self):
             pass
 
@@ -1321,12 +1321,12 @@ def test_pickling_actor_handle(ray_start_regular):
 
 def test_pickled_actor_handle_call_in_method_twice(ray_start_regular):
     @ray.remote
-    class Actor1(object):
+    class Actor1:
         def f(self):
             return 1
 
     @ray.remote
-    class Actor2(object):
+    class Actor2:
         def __init__(self, constructor):
             self.actor = constructor()
 
@@ -1345,7 +1345,7 @@ def test_register_and_get_named_actors(ray_start_regular):
     # TODO(heyucongtom): We should test this from another driver.
 
     @ray.remote
-    class Foo(object):
+    class Foo:
         def __init__(self):
             self.x = 0
 
@@ -1381,7 +1381,7 @@ def test_register_and_get_named_actors(ray_start_regular):
 
 def test_detached_actor(ray_start_regular):
     @ray.remote
-    class DetachedActor(object):
+    class DetachedActor:
         def ping(self):
             return "pong"
 
@@ -1400,7 +1400,7 @@ import ray
 ray.init(address="{}")
 
 @ray.remote
-class DetachedActor(object):
+class DetachedActor:
     def ping(self):
         return "pong"
 
@@ -1415,7 +1415,7 @@ ray.get(actor.ping.remote())
 
 def test_kill(ray_start_regular):
     @ray.remote
-    class Actor(object):
+    class Actor:
         def hang(self):
             # Never returns.
             ray.get(ray.ObjectID.from_random())
@@ -1434,7 +1434,7 @@ def test_kill(ray_start_regular):
 def test_actor_creation_task_crash(ray_start_regular):
     # Test actor death in constructor.
     @ray.remote(max_reconstructions=0)
-    class Actor(object):
+    class Actor:
         def __init__(self):
             print("crash")
             os._exit(0)
@@ -1450,7 +1450,7 @@ def test_actor_creation_task_crash(ray_start_regular):
     # Test an actor can be reconstructed successfully
     # afte it dies in its constructor.
     @ray.remote(max_reconstructions=3)
-    class ReconstructableActor(object):
+    class ReconstructableActor:
         def __init__(self):
             count = self.get_count()
             count += 1
diff --git a/python/ray/tests/test_actor_failures.py b/python/ray/tests/test_actor_failures.py
index a0d07b6f8..67dc2cecf 100644
--- a/python/ray/tests/test_actor_failures.py
+++ b/python/ray/tests/test_actor_failures.py
@@ -96,7 +96,7 @@ def test_actor_eviction(ray_start_object_store_memory):
     object_store_memory = ray_start_object_store_memory
 
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self):
             pass
 
@@ -134,7 +134,7 @@ def test_actor_reconstruction(ray_start_regular):
     """Test actor reconstruction when actor process is killed."""
 
     @ray.remote(max_reconstructions=1)
-    class ReconstructableActor(object):
+    class ReconstructableActor:
         """An actor that will be reconstructed at most once."""
 
         def __init__(self):
@@ -183,7 +183,7 @@ def test_actor_reconstruction_without_task(ray_start_regular):
     """Test a dead actor can be reconstructed without sending task to it."""
 
     @ray.remote(max_reconstructions=1)
-    class ReconstructableActor(object):
+    class ReconstructableActor:
         def __init__(self, obj_ids):
             for obj_id in obj_ids:
                 # Every time the actor gets constructed,
@@ -231,7 +231,7 @@ def test_actor_reconstruction_on_node_failure(ray_start_cluster_head):
         cluster.remove_node(node_to_remove)
 
     @ray.remote(max_reconstructions=max_reconstructions, resources={"a": 1})
-    class MyActor(object):
+    class MyActor:
         def __init__(self):
             self.value = 0
 
@@ -296,7 +296,7 @@ def test_multiple_actor_reconstruction(ray_start_cluster_head):
     ]
 
     @ray.remote(max_reconstructions=ray.ray_constants.INFINITE_RECONSTRUCTION)
-    class SlowCounter(object):
+    class SlowCounter:
         def __init__(self):
             self.x = 0
 
@@ -636,7 +636,7 @@ def test_decorated_method(ray_start_regular):
         return new_f_execution
 
     @ray.remote
-    class Actor(object):
+    class Actor:
         @method_execution_decorator
         def decorated_method(self, x):
             return x + 1
@@ -663,7 +663,7 @@ def test_ray_wait_dead_actor(ray_start_cluster):
     cluster = ray_start_cluster
 
     @ray.remote(num_cpus=1)
-    class Actor(object):
+    class Actor:
         def __init__(self):
             pass
 
@@ -707,7 +707,7 @@ def test_ray_wait_dead_actor(ray_start_cluster):
     ray.experimental.set_resource(head_node_resource, 1)
 
     @ray.remote(num_cpus=0, resources={head_node_resource: 1})
-    class ParentActor(object):
+    class ParentActor:
         def __init__(self, ping_ids):
             self.unready = ping_ids
 
diff --git a/python/ray/tests/test_actor_pool.py b/python/ray/tests/test_actor_pool.py
index a45e15a10..1822991a3 100644
--- a/python/ray/tests/test_actor_pool.py
+++ b/python/ray/tests/test_actor_pool.py
@@ -18,7 +18,7 @@ def init():
 
 def test_get_next(init):
     @ray.remote
-    class MyActor(object):
+    class MyActor:
         def __init__(self):
             pass
 
@@ -37,7 +37,7 @@ def test_get_next(init):
 
 def test_get_next_unordered(init):
     @ray.remote
-    class MyActor(object):
+    class MyActor:
         def __init__(self):
             pass
 
@@ -62,7 +62,7 @@ def test_get_next_unordered(init):
 
 def test_map(init):
     @ray.remote
-    class MyActor(object):
+    class MyActor:
         def __init__(self):
             pass
 
@@ -83,7 +83,7 @@ def test_map(init):
 
 def test_map_unordered(init):
     @ray.remote
-    class MyActor(object):
+    class MyActor:
         def __init__(self):
             pass
 
@@ -105,7 +105,7 @@ def test_map_unordered(init):
 
 def test_get_next_timeout(init):
     @ray.remote
-    class MyActor(object):
+    class MyActor:
         def __init__(self):
             pass
 
@@ -127,7 +127,7 @@ def test_get_next_timeout(init):
 
 def test_get_next_unordered_timeout(init):
     @ray.remote
-    class MyActor(object):
+    class MyActor:
         def __init__(self):
             pass
 
diff --git a/python/ray/tests/test_actor_resources.py b/python/ray/tests/test_actor_resources.py
index 2ea3bac8a..e919d6832 100644
--- a/python/ray/tests/test_actor_resources.py
+++ b/python/ray/tests/test_actor_resources.py
@@ -29,7 +29,7 @@ def test_actor_deletion_with_gpus(shutdown_only):
     # are released.
 
     @ray.remote(num_gpus=1)
-    class Actor(object):
+    class Actor:
         def getpid(self):
             return os.getpid()
 
@@ -42,7 +42,7 @@ def test_actor_deletion_with_gpus(shutdown_only):
 
 def test_actor_state(ray_start_regular):
     @ray.remote
-    class Counter(object):
+    class Counter:
         def __init__(self):
             self.value = 0
 
@@ -63,7 +63,7 @@ def test_actor_state(ray_start_regular):
 
 
 def test_actor_class_methods(ray_start_regular):
-    class Foo(object):
+    class Foo:
         x = 2
 
         @classmethod
@@ -98,7 +98,7 @@ def test_resource_assignment(shutdown_only):
         resources={"Custom": 1},
         object_store_memory=int(150 * 1024 * 1024))
 
-    class Actor(object):
+    class Actor:
         def __init__(self):
             self.resources = ray.get_resource_ids()
 
@@ -188,7 +188,7 @@ def test_actor_gpus(ray_start_cluster):
     ray.init(address=cluster.address)
 
     @ray.remote(num_gpus=1)
-    class Actor1(object):
+    class Actor1:
         def __init__(self):
             self.gpu_ids = ray.get_gpu_ids()
 
@@ -227,7 +227,7 @@ def test_actor_multiple_gpus(ray_start_cluster):
     ray.init(address=cluster.address)
 
     @ray.remote(num_gpus=2)
-    class Actor1(object):
+    class Actor1:
         def __init__(self):
             self.gpu_ids = ray.get_gpu_ids()
 
@@ -259,7 +259,7 @@ def test_actor_multiple_gpus(ray_start_cluster):
 
     # We should be able to create more actors that use only a single GPU.
     @ray.remote(num_gpus=1)
-    class Actor2(object):
+    class Actor2:
         def __init__(self):
             self.gpu_ids = ray.get_gpu_ids()
 
@@ -297,7 +297,7 @@ def test_actor_different_numbers_of_gpus(ray_start_cluster):
     ray.init(address=cluster.address)
 
     @ray.remote(num_gpus=1)
-    class Actor1(object):
+    class Actor1:
         def __init__(self):
             self.gpu_ids = ray.get_gpu_ids()
 
@@ -343,7 +343,7 @@ def test_actor_multiple_gpus_from_multiple_tasks(ray_start_cluster):
     @ray.remote
     def create_actors(i, n):
         @ray.remote(num_gpus=1)
-        class Actor(object):
+        class Actor:
             def __init__(self, i, j):
                 self.gpu_ids = ray.get_gpu_ids()
 
@@ -389,7 +389,7 @@ def test_actor_multiple_gpus_from_multiple_tasks(ray_start_cluster):
         assert len(set(gpus_in_use[node_name])) == num_gpus_per_raylet
 
     @ray.remote(num_gpus=1)
-    class Actor(object):
+    class Actor:
         def __init__(self):
             self.gpu_ids = ray.get_gpu_ids()
 
@@ -452,7 +452,7 @@ def test_actors_and_tasks_with_gpus(ray_start_cluster):
                 [t1, t2])
 
     @ray.remote(num_gpus=1)
-    class Actor1(object):
+    class Actor1:
         def __init__(self):
             self.gpu_ids = ray.get_gpu_ids()
             assert len(self.gpu_ids) == 1
@@ -528,7 +528,7 @@ def test_actors_and_tasks_with_gpus_version_two(shutdown_only):
     # for a long time in order to make sure the GPU is not released
     # prematurely.
     @ray.remote
-    class RecordGPUs(object):
+    class RecordGPUs:
         def __init__(self):
             self.gpu_ids_seen = []
             self.num_calls = 0
@@ -550,7 +550,7 @@ def test_actors_and_tasks_with_gpus_version_two(shutdown_only):
         time.sleep(1000)
 
     @ray.remote(num_gpus=1)
-    class Actor(object):
+    class Actor:
         def __init__(self, record_gpu_actor):
             self.gpu_ids = ray.get_gpu_ids()
             assert len(self.gpu_ids) == 1
@@ -592,7 +592,7 @@ def test_blocking_actor_task(shutdown_only):
         return 1
 
     @ray.remote
-    class Foo(object):
+    class Foo:
         def __init__(self):
             pass
 
@@ -605,7 +605,7 @@ def test_blocking_actor_task(shutdown_only):
     ray.get(actor.blocking_method.remote())
 
     @ray.remote(num_cpus=1)
-    class CPUFoo(object):
+    class CPUFoo:
         def __init__(self):
             pass
 
@@ -621,7 +621,7 @@ def test_blocking_actor_task(shutdown_only):
     assert remaining_ids == [x_id]
 
     @ray.remote(num_gpus=1)
-    class GPUFoo(object):
+    class GPUFoo:
         def __init__(self):
             pass
 
@@ -639,13 +639,13 @@ def test_blocking_actor_task(shutdown_only):
 def test_lifetime_and_transient_resources(ray_start_regular):
     # This actor acquires resources only when running methods.
     @ray.remote
-    class Actor1(object):
+    class Actor1:
         def method(self):
             pass
 
     # This actor acquires resources for its lifetime.
     @ray.remote(num_cpus=1)
-    class Actor2(object):
+    class Actor2:
         def method(self):
             pass
 
@@ -666,12 +666,12 @@ def test_custom_label_placement(ray_start_cluster):
     ray.init(address=cluster.address)
 
     @ray.remote(resources={"CustomResource1": 1})
-    class ResourceActor1(object):
+    class ResourceActor1:
         def get_location(self):
             return ray.worker.global_worker.node.unique_id
 
     @ray.remote(resources={"CustomResource2": 1})
-    class ResourceActor2(object):
+    class ResourceActor2:
         def get_location(self):
             return ray.worker.global_worker.node.unique_id
 
@@ -692,12 +692,12 @@ def test_creating_more_actors_than_resources(shutdown_only):
     ray.init(num_cpus=10, num_gpus=2, resources={"CustomResource1": 1})
 
     @ray.remote(num_gpus=1)
-    class ResourceActor1(object):
+    class ResourceActor1:
         def method(self):
             return ray.get_gpu_ids()[0]
 
     @ray.remote(resources={"CustomResource1": 1})
-    class ResourceActor2(object):
+    class ResourceActor2:
         def method(self):
             pass
 
diff --git a/python/ray/tests/test_advanced.py b/python/ray/tests/test_advanced.py
index 43074c60d..262c6f8a4 100644
--- a/python/ray/tests/test_advanced.py
+++ b/python/ray/tests/test_advanced.py
@@ -373,7 +373,7 @@ def test_multithreading(ray_start_2_cpus):
         """Test using Ray api in multiple threads."""
 
         @ray.remote
-        class Echo(object):
+        class Echo:
             def echo(self, value):
                 return value
 
@@ -433,7 +433,7 @@ def test_multithreading(ray_start_2_cpus):
 
     # Test actor that runs background threads.
     @ray.remote
-    class MultithreadedActor(object):
+    class MultithreadedActor:
         def __init__(self):
             self.lock = threading.Lock()
             self.thread_results = []
@@ -504,7 +504,7 @@ def test_free_objects_multi_node(ray_start_cluster):
             _internal_config=config)
     ray.init(address=cluster.address)
 
-    class RawActor(object):
+    class RawActor:
         def get(self):
             return ray.worker.global_worker.node.unique_id
 
@@ -633,7 +633,7 @@ def test_local_mode(shutdown_only):
     # Test actors in LOCAL_MODE.
 
     @ray.remote
-    class LocalModeTestClass(object):
+    class LocalModeTestClass:
         def __init__(self, array):
             self.array = array
 
@@ -718,7 +718,7 @@ def test_local_mode(shutdown_only):
     # Check that Actors are not overwritten by remote calls from different
     # classes.
     @ray.remote
-    class RemoteActor1(object):
+    class RemoteActor1:
         def __init__(self):
             pass
 
@@ -726,7 +726,7 @@ def test_local_mode(shutdown_only):
             return 0
 
     @ray.remote
-    class RemoteActor2(object):
+    class RemoteActor2:
         def __init__(self):
             pass
 
@@ -756,7 +756,7 @@ def test_wait_makes_object_local(ray_start_cluster):
     ray.init(address=cluster.address)
 
     @ray.remote
-    class Foo(object):
+    class Foo:
         def method(self):
             return np.zeros(1024 * 1024)
 
diff --git a/python/ray/tests/test_advanced_2.py b/python/ray/tests/test_advanced_2.py
index 03c0b11a8..a9e50bd1f 100644
--- a/python/ray/tests/test_advanced_2.py
+++ b/python/ray/tests/test_advanced_2.py
@@ -191,7 +191,7 @@ def test_gpu_ids(shutdown_only):
     # Test that actors have CUDA_VISIBLE_DEVICES set properly.
 
     @ray.remote
-    class Actor0(object):
+    class Actor0:
         def __init__(self):
             gpu_ids = ray.get_gpu_ids()
             assert len(gpu_ids) == 0
@@ -208,7 +208,7 @@ def test_gpu_ids(shutdown_only):
             return self.x
 
     @ray.remote(num_gpus=1)
-    class Actor1(object):
+    class Actor1:
         def __init__(self):
             gpu_ids = ray.get_gpu_ids()
             assert len(gpu_ids) == 1
@@ -243,7 +243,7 @@ def test_zero_cpus(shutdown_only):
 
     # We should be able to create an actor that requires 0 CPU resources.
     @ray.remote(num_cpus=0)
-    class Actor(object):
+    class Actor:
         def method(self):
             pass
 
@@ -261,7 +261,7 @@ def test_zero_cpus_actor(ray_start_cluster):
     node_id = ray.worker.global_worker.node.unique_id
 
     @ray.remote
-    class Foo(object):
+    class Foo:
         def method(self):
             return ray.worker.global_worker.node.unique_id
 
@@ -274,7 +274,7 @@ def test_fractional_resources(shutdown_only):
     ray.init(num_cpus=6, num_gpus=3, resources={"Custom": 1})
 
     @ray.remote(num_gpus=0.5)
-    class Foo1(object):
+    class Foo1:
         def method(self):
             gpu_ids = ray.get_gpu_ids()
             assert len(gpu_ids) == 1
@@ -287,7 +287,7 @@ def test_fractional_resources(shutdown_only):
     del foos
 
     @ray.remote
-    class Foo2(object):
+    class Foo2:
         def method(self):
             pass
 
diff --git a/python/ray/tests/test_advanced_3.py b/python/ray/tests/test_advanced_3.py
index 52089844c..f7d56f735 100644
--- a/python/ray/tests/test_advanced_3.py
+++ b/python/ray/tests/test_advanced_3.py
@@ -260,7 +260,7 @@ def test_global_state_task_object_api(shutdown_only):
 
 # TODO(rkn): Pytest actually has tools for capturing stdout and stderr, so we
 # should use those, but they seem to conflict with Ray's use of faulthandler.
-class CaptureOutputAndError(object):
+class CaptureOutputAndError:
     """Capture stdout and stderr of some span.
 
     This can be used as follows.
@@ -434,7 +434,7 @@ def test_wait_reconstruction(shutdown_only):
 
 def test_ray_setproctitle(ray_start_2_cpus):
     @ray.remote
-    class UniqueName(object):
+    class UniqueName:
         def __init__(self):
             assert setproctitle.getproctitle() == "ray::UniqueName.__init__()"
 
@@ -667,7 +667,7 @@ def test_export_after_shutdown(ray_start_regular):
         pass
 
     @ray.remote
-    class Actor(object):
+    class Actor:
         def method(self):
             pass
 
@@ -731,7 +731,7 @@ def test_move_log_files_to_old(shutdown_only):
     logs_dir = os.path.join(info["session_dir"], "logs")
 
     @ray.remote
-    class Actor(object):
+    class Actor:
         def f(self):
             print("function f finished")
 
diff --git a/python/ray/tests/test_autoscaler.py b/python/ray/tests/test_autoscaler.py
index 46108f645..ff23e70f9 100644
--- a/python/ray/tests/test_autoscaler.py
+++ b/python/ray/tests/test_autoscaler.py
@@ -21,7 +21,7 @@ from ray.test_utils import RayTestTimeoutException
 import pytest
 
 
-class MockNode(object):
+class MockNode:
     def __init__(self, node_id, tags):
         self.node_id = node_id
         self.state = "pending"
@@ -36,7 +36,7 @@ class MockNode(object):
         return True
 
 
-class MockProcessRunner(object):
+class MockProcessRunner:
     def __init__(self, fail_cmds=[]):
         self.calls = []
         self.fail_cmds = fail_cmds
diff --git a/python/ray/tests/test_basic.py b/python/ray/tests/test_basic.py
index 8a875209b..bc190cb78 100644
--- a/python/ray/tests/test_basic.py
+++ b/python/ray/tests/test_basic.py
@@ -217,7 +217,7 @@ def complex_serialization(use_pickle):
         },
     ]
 
-    class Foo(object):
+    class Foo:
         def __init__(self, value=0):
             self.value = value
 
@@ -227,12 +227,12 @@ def complex_serialization(use_pickle):
         def __eq__(self, other):
             return other.value == self.value
 
-    class Bar(object):
+    class Bar:
         def __init__(self):
             for i, val in enumerate(PRIMITIVE_OBJECTS + COMPLEX_OBJECTS):
                 setattr(self, "field{}".format(i), val)
 
-    class Baz(object):
+    class Baz:
         def __init__(self):
             self.foo = Foo()
             self.bar = Bar()
@@ -240,7 +240,7 @@ def complex_serialization(use_pickle):
         def method(self, arg):
             pass
 
-    class Qux(object):
+    class Qux:
         def __init__(self):
             self.objs = [Foo(), Bar(), Baz()]
 
@@ -273,7 +273,7 @@ def complex_serialization(use_pickle):
 
         CUSTOM_OBJECTS.append(DataClass0(number=3))
 
-        class CustomClass(object):
+        class CustomClass:
             def __init__(self, value):
                 self.value = value
 
@@ -385,7 +385,7 @@ def test_nested_functions(ray_start_regular):
 
 
 def test_ray_recursive_objects(ray_start_regular):
-    class ClassA(object):
+    class ClassA:
         pass
 
     # Make a list that contains itself.
@@ -440,7 +440,7 @@ def test_passing_arguments_by_value_out_of_the_box(ray_start_regular):
     assert ray.get(f.remote(float)) == float
     assert ray.get(f.remote(str)) == str
 
-    class Foo(object):
+    class Foo:
         def __init__(self):
             pass
 
@@ -453,7 +453,7 @@ def test_putting_object_that_closes_over_object_id(ray_start_regular):
     # This test is here to prevent a regression of
     # https://github.com/ray-project/ray/issues/1317.
 
-    class Foo(object):
+    class Foo:
         def __init__(self):
             self.val = ray.put(0)
 
@@ -493,7 +493,7 @@ def test_put_get(shutdown_only):
 
 
 def custom_serializers():
-    class Foo(object):
+    class Foo:
         def __init__(self):
             self.x = 3
 
@@ -508,7 +508,7 @@ def custom_serializers():
 
     assert ray.get(ray.put(Foo())) == ((3, "string1", Foo.__name__), "string2")
 
-    class Bar(object):
+    class Bar:
         def __init__(self):
             self.x = 3
 
@@ -530,7 +530,7 @@ def test_custom_serializers_with_pickle(shutdown_only):
     ray.init(use_pickle=True)
     custom_serializers()
 
-    class Foo(object):
+    class Foo:
         def __init__(self):
             self.x = 4
 
@@ -566,7 +566,7 @@ def test_serialization_final_fallback(ray_start_regular):
 def test_register_class(ray_start_2_cpus):
     # Check that putting an object of a class that has not been registered
     # throws an exception.
-    class TempClass(object):
+    class TempClass:
         pass
 
     ray.get(ray.put(TempClass()))
@@ -576,7 +576,7 @@ def test_register_class(ray_start_2_cpus):
     def f(x):
         return x
 
-    class Foo(object):
+    class Foo:
         def __init__(self, value=0):
             self.value = value
 
@@ -597,14 +597,14 @@ def test_register_class(ray_start_2_cpus):
     # Instead, we do this:
     assert regex.pattern == new_regex.pattern
 
-    class TempClass1(object):
+    class TempClass1:
         def __init__(self):
             self.value = 1
 
     # Test returning custom classes created on workers.
     @ray.remote
     def g():
-        class TempClass2(object):
+        class TempClass2:
             def __init__(self):
                 self.value = 2
 
@@ -616,7 +616,7 @@ def test_register_class(ray_start_2_cpus):
 
     # Test exporting custom class definitions from one worker to another
     # when the worker is blocked in a get.
-    class NewTempClass(object):
+    class NewTempClass:
         def __init__(self, value):
             self.value = value
 
@@ -633,19 +633,19 @@ def test_register_class(ray_start_2_cpus):
     # Test registering multiple classes with the same name.
     @ray.remote(num_return_vals=3)
     def j():
-        class Class0(object):
+        class Class0:
             def method0(self):
                 pass
 
         c0 = Class0()
 
-        class Class0(object):
+        class Class0:
             def method1(self):
                 pass
 
         c1 = Class0()
 
-        class Class0(object):
+        class Class0:
             def method2(self):
                 pass
 
@@ -672,19 +672,19 @@ def test_register_class(ray_start_2_cpus):
 
     @ray.remote
     def k():
-        class Class0(object):
+        class Class0:
             def method0(self):
                 pass
 
         c0 = Class0()
 
-        class Class0(object):
+        class Class0:
             def method1(self):
                 pass
 
         c1 = Class0()
 
-        class Class0(object):
+        class Class0:
             def method2(self):
                 pass
 
@@ -804,7 +804,7 @@ def test_args_starkwargs(ray_start_regular):
     def starkwargs(a, b, **kwargs):
         return a, b, kwargs
 
-    class TestActor(object):
+    class TestActor:
         def starkwargs(self, a, b, **kwargs):
             return a, b, kwargs
 
@@ -841,7 +841,7 @@ def test_args_named_and_star(ray_start_regular):
     def hello(a, x="hello", **kwargs):
         return a, x, kwargs
 
-    class TestActor(object):
+    class TestActor:
         def hello(self, a, x="hello", **kwargs):
             return a, x, kwargs
 
@@ -884,7 +884,7 @@ def test_args_stars_after(ray_start_regular):
     def star_args_after(a="hello", b="heo", *args, **kwargs):
         return a, b, args, kwargs
 
-    class TestActor(object):
+    class TestActor:
         def star_args_after(self, a="hello", b="heo", *args, **kwargs):
             return a, b, args, kwargs
 
@@ -1089,7 +1089,7 @@ def test_submit_api(shutdown_only):
     assert len(remaining_ids) == 1
 
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self, x, y=0):
             self.x = x
             self.y = y
@@ -1101,7 +1101,7 @@ def test_submit_api(shutdown_only):
             return ray.get_gpu_ids()
 
     @ray.remote
-    class Actor2(object):
+    class Actor2:
         def __init__(self):
             pass
 
@@ -1250,7 +1250,7 @@ def test_direct_call_simple(ray_start_cluster):
 # https://github.com/ray-project/ray/issues/6329
 def test_call_actors_indirect_through_tasks(ray_start_regular):
     @ray.remote
-    class Counter(object):
+    class Counter:
         def __init__(self, value):
             self.value = int(value)
 
@@ -1302,7 +1302,7 @@ def test_direct_call_matrix(shutdown_only):
     ray.init(object_store_memory=1000 * 1024 * 1024)
 
     @ray.remote
-    class Actor(object):
+    class Actor:
         def small_value(self):
             return 0
 
@@ -1390,7 +1390,7 @@ def test_direct_inline_arg_memory_corruption(ray_start_regular):
         return np.zeros(1000, dtype=np.uint8)
 
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self):
             self.z = []
 
@@ -1407,7 +1407,7 @@ def test_direct_inline_arg_memory_corruption(ray_start_regular):
 
 def test_direct_actor_enabled(ray_start_regular):
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self):
             pass
 
@@ -1430,7 +1430,7 @@ def test_direct_actor_order(shutdown_only):
         return 0
 
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self):
             self.count = 0
 
@@ -1448,7 +1448,7 @@ def test_direct_actor_order(shutdown_only):
 
 def test_direct_actor_large_objects(ray_start_regular):
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self):
             pass
 
@@ -1467,7 +1467,7 @@ def test_direct_actor_large_objects(ray_start_regular):
 
 def test_direct_actor_pass_by_ref(ray_start_regular):
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self):
             pass
 
@@ -1497,7 +1497,7 @@ def test_direct_actor_pass_by_ref_order_optimization(shutdown_only):
     ray.init(num_cpus=4)
 
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self):
             pass
 
@@ -1532,7 +1532,7 @@ def test_direct_actor_pass_by_ref_order_optimization(shutdown_only):
 
 def test_direct_actor_recursive(ray_start_regular):
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self, delegate=None):
             self.delegate = delegate
 
@@ -1555,7 +1555,7 @@ def test_direct_actor_recursive(ray_start_regular):
 
 def test_direct_actor_concurrent(ray_start_regular):
     @ray.remote
-    class Batcher(object):
+    class Batcher:
         def __init__(self):
             self.batch = []
             self.event = threading.Event()
diff --git a/python/ray/tests/test_component_failures_3.py b/python/ray/tests/test_component_failures_3.py
index 029fdcd2d..e8164881b 100644
--- a/python/ray/tests/test_component_failures_3.py
+++ b/python/ray/tests/test_component_failures_3.py
@@ -25,7 +25,7 @@ def test_actor_creation_node_failure(ray_start_cluster):
     cluster = ray_start_cluster
 
     @ray.remote
-    class Child(object):
+    class Child:
         def __init__(self, death_probability):
             self.death_probability = death_probability
 
diff --git a/python/ray/tests/test_dynres.py b/python/ray/tests/test_dynres.py
index 50d782c37..3dbdc0594 100644
--- a/python/ray/tests/test_dynres.py
+++ b/python/ray/tests/test_dynres.py
@@ -580,7 +580,7 @@ def test_release_cpus_when_actor_creation_task_blocking(shutdown_only):
         return 100
 
     @ray.remote(num_cpus=1)
-    class A(object):
+    class A:
         def __init__(self):
             self.num = ray.get(get_100.remote())
 
diff --git a/python/ray/tests/test_failure.py b/python/ray/tests/test_failure.py
index 680d27ed8..0669e0a98 100644
--- a/python/ray/tests/test_failure.py
+++ b/python/ray/tests/test_failure.py
@@ -167,7 +167,7 @@ def temporary_helper_function():
     # Define an actor that closes over this temporary module. This should
     # fail when it is unpickled.
     @ray.remote
-    class Foo(object):
+    class Foo:
         def __init__(self, arg1, arg2=3):
             self.x = module.temporary_python_file()
 
@@ -213,7 +213,7 @@ def test_failed_actor_init(ray_start_regular):
     error_message2 = "actor method failed"
 
     @ray.remote
-    class FailedActor(object):
+    class FailedActor:
         def __init__(self):
             raise Exception(error_message1)
 
@@ -240,7 +240,7 @@ def test_failed_actor_method(ray_start_regular):
     error_message2 = "actor method failed"
 
     @ray.remote
-    class FailedActor(object):
+    class FailedActor:
         def __init__(self):
             pass
 
@@ -259,7 +259,7 @@ def test_failed_actor_method(ray_start_regular):
 
 def test_incorrect_method_calls(ray_start_regular):
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self, missing_variable_name):
             pass
 
@@ -325,7 +325,7 @@ def test_worker_dying(ray_start_regular):
 
 def test_actor_worker_dying(ray_start_regular):
     @ray.remote
-    class Actor(object):
+    class Actor:
         def kill(self):
             eval("exit()")
 
@@ -344,7 +344,7 @@ def test_actor_worker_dying(ray_start_regular):
 
 def test_actor_worker_dying_future_tasks(ray_start_regular):
     @ray.remote(max_reconstructions=0)
-    class Actor(object):
+    class Actor:
         def getpid(self):
             return os.getpid()
 
@@ -366,7 +366,7 @@ def test_actor_worker_dying_future_tasks(ray_start_regular):
 
 def test_actor_worker_dying_nothing_in_progress(ray_start_regular):
     @ray.remote(max_reconstructions=0)
-    class Actor(object):
+    class Actor:
         def getpid(self):
             return os.getpid()
 
@@ -381,7 +381,7 @@ def test_actor_worker_dying_nothing_in_progress(ray_start_regular):
 
 def test_actor_scope_or_intentionally_killed_message(ray_start_regular):
     @ray.remote
-    class Actor(object):
+    class Actor:
         pass
 
     a = Actor.remote()
@@ -532,7 +532,7 @@ def test_export_large_objects(ray_start_regular):
     wait_for_errors(ray_constants.PICKLING_LARGE_OBJECT_PUSH_ERROR, 1)
 
     @ray.remote
-    class Foo(object):
+    class Foo:
         def __init__(self):
             large_object
 
@@ -548,7 +548,7 @@ def test_warning_for_resource_deadlock(shutdown_only):
     ray.init(num_cpus=1)
 
     @ray.remote(num_cpus=1)
-    class Foo(object):
+    class Foo:
         def f(self):
             return 0
 
@@ -572,7 +572,7 @@ def test_warning_for_infeasible_tasks(ray_start_regular):
         pass
 
     @ray.remote(resources={"Custom": 1})
-    class Foo(object):
+    class Foo:
         pass
 
     # This task is infeasible.
@@ -592,7 +592,7 @@ def test_warning_for_infeasible_zero_cpu_actor(shutdown_only):
     ray.init(num_cpus=0)
 
     @ray.remote
-    class Foo(object):
+    class Foo:
         pass
 
     # The actor creation should be infeasible.
@@ -607,7 +607,7 @@ def test_warning_for_too_many_actors(shutdown_only):
     ray.init(num_cpus=num_cpus)
 
     @ray.remote
-    class Foo(object):
+    class Foo:
         def __init__(self):
             time.sleep(1000)
 
@@ -689,7 +689,7 @@ def test_warning_for_many_duplicate_remote_functions_and_actors(shutdown_only):
         # Require a GPU so that the actor is never actually created and we
         # don't spawn an unreasonable number of processes.
         @ray.remote(num_gpus=1)
-        class Foo(object):
+        class Foo:
             pass
 
         Foo.remote()
@@ -850,7 +850,7 @@ def test_connect_with_disconnected_node(shutdown_only):
 @pytest.mark.parametrize("num_actors", [1, 2, 5])
 def test_parallel_actor_fill_plasma_retry(ray_start_cluster_head, num_actors):
     @ray.remote
-    class LargeMemoryActor(object):
+    class LargeMemoryActor:
         def some_expensive_task(self):
             return np.zeros(10**8 // 2, dtype=np.uint8)
 
@@ -869,7 +869,7 @@ def test_parallel_actor_fill_plasma_retry(ray_start_cluster_head, num_actors):
     indirect=True)
 def test_fill_object_store_exception(ray_start_cluster_head):
     @ray.remote
-    class LargeMemoryActor(object):
+    class LargeMemoryActor:
         def some_expensive_task(self):
             return np.zeros(10**8 + 2, dtype=np.uint8)
 
diff --git a/python/ray/tests/test_garbage_collection.py b/python/ray/tests/test_garbage_collection.py
index bc2e20565..b6ce12501 100644
--- a/python/ray/tests/test_garbage_collection.py
+++ b/python/ray/tests/test_garbage_collection.py
@@ -30,7 +30,7 @@ def test_basic_gc(shutdown_only):
         return np.random.shuffle(input)
 
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self):
             # Hold a long-lived reference to a ray.put object. This should not
             # be garbage collected while the actor is alive.
diff --git a/python/ray/tests/test_memory_limits.py b/python/ray/tests/test_memory_limits.py
index c19021a6a..03a1c3861 100644
--- a/python/ray/tests/test_memory_limits.py
+++ b/python/ray/tests/test_memory_limits.py
@@ -10,7 +10,7 @@ OBJECT_TOO_LARGE = ray.exceptions.ObjectStoreFullError
 
 
 @ray.remote
-class LightActor(object):
+class LightActor:
     def __init__(self):
         pass
 
@@ -19,7 +19,7 @@ class LightActor(object):
 
 
 @ray.remote
-class GreedyActor(object):
+class GreedyActor:
     def __init__(self):
         pass
 
diff --git a/python/ray/tests/test_memory_scheduling.py b/python/ray/tests/test_memory_scheduling.py
index ebcebd54b..7e067cb32 100644
--- a/python/ray/tests/test_memory_scheduling.py
+++ b/python/ray/tests/test_memory_scheduling.py
@@ -9,7 +9,7 @@ MB = 1024 * 1024
 
 
 @ray.remote(memory=100 * MB)
-class Actor(object):
+class Actor:
     def __init__(self):
         pass
 
@@ -18,7 +18,7 @@ class Actor(object):
 
 
 @ray.remote(object_store_memory=100 * MB)
-class Actor2(object):
+class Actor2:
     def __init__(self):
         pass
 
diff --git a/python/ray/tests/test_metrics.py b/python/ray/tests/test_metrics.py
index 473a41fb4..46c33917d 100644
--- a/python/ray/tests/test_metrics.py
+++ b/python/ray/tests/test_metrics.py
@@ -48,7 +48,7 @@ def test_worker_stats(shutdown_only):
         return os.getpid()
 
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self):
             pass
 
@@ -118,7 +118,7 @@ def test_raylet_info_endpoint(shutdown_only):
     addresses = ray.init(include_webui=True, num_cpus=6)
 
     @ray.remote(num_cpus=1)
-    class ActorA(object):
+    class ActorA:
         def __init__(self):
             pass
 
@@ -126,12 +126,12 @@ def test_raylet_info_endpoint(shutdown_only):
             return os.getpid()
 
     @ray.remote(resources={"CustomResource": 1})
-    class ActorB(object):
+    class ActorB:
         def __init__(self):
             pass
 
     @ray.remote(num_cpus=2)
-    class ActorC(object):
+    class ActorC:
         def __init__(self):
             self.children = [ActorA.remote(), ActorB.remote()]
 
diff --git a/python/ray/tests/test_mini.py b/python/ray/tests/test_mini.py
index b4e630080..605b91056 100644
--- a/python/ray/tests/test_mini.py
+++ b/python/ray/tests/test_mini.py
@@ -53,7 +53,7 @@ def test_put_api(ray_start_regular):
 
 def test_actor_api(ray_start_regular):
     @ray.remote
-    class Foo(object):
+    class Foo:
         def __init__(self, val):
             self.x = val
 
diff --git a/python/ray/tests/test_multi_node.py b/python/ray/tests/test_multi_node.py
index 65506d664..4eb83a529 100644
--- a/python/ray/tests/test_multi_node.py
+++ b/python/ray/tests/test_multi_node.py
@@ -143,7 +143,7 @@ def test_driver_exiting_quickly(call_ray_start):
 import ray
 ray.init(address="{}")
 @ray.remote
-class Foo(object):
+class Foo:
     def __init__(self):
         pass
 Foo.remote()
@@ -185,7 +185,7 @@ import ray
 import time
 ray.init(address="{}")
 @ray.remote
-class Counter(object):
+class Counter:
     def __init__(self):
         self.count = 0
     def increment(self):
@@ -236,7 +236,7 @@ import time
 log_message = "{}"
 
 @ray.remote
-class Actor(object):
+class Actor:
     def log(self):
         print(log_message)
 
@@ -278,7 +278,7 @@ def g(duration):
     time.sleep(duration)
 
 @ray.remote(num_gpus=1)
-class Foo(object):
+class Foo:
     def __init__(self):
         pass
 
diff --git a/python/ray/tests/test_multi_node_2.py b/python/ray/tests/test_multi_node_2.py
index 4608e490b..fa986e1d8 100644
--- a/python/ray/tests/test_multi_node_2.py
+++ b/python/ray/tests/test_multi_node_2.py
@@ -138,7 +138,7 @@ def test_heartbeats_single(ray_start_cluster_head):
     ray.get(work_handle)
 
     @ray.remote
-    class Actor(object):
+    class Actor:
         def work(self, timeout):
             time.sleep(timeout)
             return True
diff --git a/python/ray/tests/test_multinode_failures_2.py b/python/ray/tests/test_multinode_failures_2.py
index 879aa9231..f03fee580 100644
--- a/python/ray/tests/test_multinode_failures_2.py
+++ b/python/ray/tests/test_multinode_failures_2.py
@@ -94,7 +94,7 @@ def test_actor_creation_node_failure(ray_start_cluster):
     cluster = ray_start_cluster
 
     @ray.remote
-    class Child(object):
+    class Child:
         def __init__(self, death_probability):
             self.death_probability = death_probability
 
diff --git a/python/ray/tests/test_object_manager.py b/python/ray/tests/test_object_manager.py
index f6eee16e1..186189f20 100644
--- a/python/ray/tests/test_object_manager.py
+++ b/python/ray/tests/test_object_manager.py
@@ -133,7 +133,7 @@ def test_actor_broadcast(ray_start_cluster_with_resource):
     cluster, num_nodes = ray_start_cluster_with_resource
 
     @ray.remote
-    class Actor(object):
+    class Actor:
         def ready(self):
             pass
 
diff --git a/python/ray/tests/test_ray_init.py b/python/ray/tests/test_ray_init.py
index 8ce3f07d8..cf3ecf4cd 100644
--- a/python/ray/tests/test_ray_init.py
+++ b/python/ray/tests/test_ray_init.py
@@ -18,7 +18,7 @@ def password():
     return random_bytes.encode("hex")  # Python 2
 
 
-class TestRedisPassword(object):
+class TestRedisPassword:
     @pytest.mark.skipif(
         os.environ.get("RAY_USE_NEW_GCS") == "on",
         reason="New GCS API doesn't support Redis authentication yet.")
diff --git a/python/ray/tests/test_signal.py b/python/ray/tests/test_signal.py
index 078fe5a3f..af41c5126 100644
--- a/python/ray/tests/test_signal.py
+++ b/python/ray/tests/test_signal.py
@@ -42,7 +42,7 @@ def test_send_signal_from_actor_to_driver(ray_start_regular):
     # Send several signals from an actor, and receive them in the driver.
 
     @ray.remote
-    class ActorSendSignal(object):
+    class ActorSendSignal:
         def __init__(self):
             pass
 
@@ -66,7 +66,7 @@ def test_send_signals_from_actor_to_driver(ray_start_regular):
     # these signals in the driver.
 
     @ray.remote
-    class ActorSendSignals(object):
+    class ActorSendSignals:
         def __init__(self):
             pass
 
@@ -123,7 +123,7 @@ def test_actor_crash(ray_start_regular):
     # of a method that failed.
 
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self):
             pass
 
@@ -145,7 +145,7 @@ def test_actor_crash_init(ray_start_regular):
     # Get an error when an actor's __init__ failed.
 
     @ray.remote
-    class ActorCrashInit(object):
+    class ActorCrashInit:
         def __init__(self):
             raise Exception("exception message")
 
@@ -165,7 +165,7 @@ def test_actor_crash_init2(ray_start_regular):
     # of the actor.
 
     @ray.remote
-    class ActorCrashInit(object):
+    class ActorCrashInit:
         def __init__(self):
             raise Exception("exception message")
 
@@ -188,7 +188,7 @@ def test_actor_crash_init3(ray_start_regular):
     # another method of the actor is invoked.
 
     @ray.remote
-    class ActorCrashInit(object):
+    class ActorCrashInit:
         def __init__(self):
             raise Exception("exception message")
 
@@ -209,7 +209,7 @@ def test_send_signals_from_actor_to_actor(ray_start_regular):
     # these signals in another actor.
 
     @ray.remote
-    class ActorSendSignals(object):
+    class ActorSendSignals:
         def __init__(self):
             pass
 
@@ -218,7 +218,7 @@ def test_send_signals_from_actor_to_actor(ray_start_regular):
                 signal.send(UserSignal(value + str(i)))
 
     @ray.remote
-    class ActorGetSignalsAll(object):
+    class ActorGetSignalsAll:
         def __init__(self):
             self.received_signals = []
 
@@ -257,7 +257,7 @@ def test_forget(ray_start_regular):
     # actor. Then show that the driver only gets the last "count" signals.
 
     @ray.remote
-    class ActorSendSignals(object):
+    class ActorSendSignals:
         def __init__(self):
             pass
 
@@ -278,7 +278,7 @@ def test_forget(ray_start_regular):
 def test_signal_on_node_failure(two_node_cluster):
     """Test actor checkpointing on a remote node."""
 
-    class ActorSignal(object):
+    class ActorSignal:
         def __init__(self):
             pass
 
diff --git a/python/ray/tests/test_stress_sharded.py b/python/ray/tests/test_stress_sharded.py
index ec738a385..a55ba2503 100644
--- a/python/ray/tests/test_stress_sharded.py
+++ b/python/ray/tests/test_stress_sharded.py
@@ -48,7 +48,7 @@ def test_submitting_many_tasks(ray_start_sharded):
 
 def test_submitting_many_actors_to_one(ray_start_sharded):
     @ray.remote
-    class Actor(object):
+    class Actor:
         def __init__(self):
             pass
 
@@ -56,7 +56,7 @@ def test_submitting_many_actors_to_one(ray_start_sharded):
             return
 
     @ray.remote
-    class Worker(object):
+    class Worker:
         def __init__(self, actor):
             self.actor = actor
 
diff --git a/python/ray/tests/test_tensorflow.py b/python/ray/tests/test_tensorflow.py
index 9e15f8616..56f02fcc4 100644
--- a/python/ray/tests/test_tensorflow.py
+++ b/python/ray/tests/test_tensorflow.py
@@ -22,7 +22,7 @@ def make_linear_network(w_name=None, b_name=None):
             tf.global_variables_initializer(), x_data, y_data)
 
 
-class LossActor(object):
+class LossActor:
     def __init__(self, use_loss=True):
         # Uses a separate graph for each network.
         with tf.Graph().as_default():
@@ -45,7 +45,7 @@ class LossActor(object):
         return self.values[0].get_weights()
 
 
-class NetActor(object):
+class NetActor:
     def __init__(self):
         # Uses a separate graph for each network.
         with tf.Graph().as_default():
@@ -67,7 +67,7 @@ class NetActor(object):
         return self.values[0].get_weights()
 
 
-class TrainActor(object):
+class TrainActor:
     def __init__(self):
         # Almost the same as above, but now returns the placeholders and
         # gradient.
diff --git a/python/ray/tune/analysis/experiment_analysis.py b/python/ray/tune/analysis/experiment_analysis.py
index 0ebe99192..794bd5d64 100644
--- a/python/ray/tune/analysis/experiment_analysis.py
+++ b/python/ray/tune/analysis/experiment_analysis.py
@@ -17,7 +17,7 @@ from ray.tune.result import EXPR_PROGRESS_FILE, EXPR_PARAM_FILE, CONFIG_PREFIX
 logger = logging.getLogger(__name__)
 
 
-class Analysis(object):
+class Analysis:
     """Analyze all results from a directory of experiments."""
 
     def __init__(self, experiment_dir):
diff --git a/python/ray/tune/automl/search_space.py b/python/ray/tune/automl/search_space.py
index 65db657e8..2f7743c69 100644
--- a/python/ray/tune/automl/search_space.py
+++ b/python/ray/tune/automl/search_space.py
@@ -11,7 +11,7 @@ from ray.tune import grid_search
 logger = logging.getLogger(__name__)
 
 
-class ParameterSpace(object):
+class ParameterSpace:
     """Base class of a single parameter's search space.
     """
 
@@ -116,7 +116,7 @@ class ContinuousSpace(ParameterSpace):
                                                  self.end)
 
 
-class SearchSpace(object):
+class SearchSpace:
     """Collection of ``ParameterSpace``, a.k.a <name, space> pair.
 
     It's supposed to be used with a fixed experiment config, which
diff --git a/python/ray/tune/automlboard/backend/collector.py b/python/ray/tune/automlboard/backend/collector.py
index dd87df1a4..f9b1bd488 100644
--- a/python/ray/tune/automlboard/backend/collector.py
+++ b/python/ray/tune/automlboard/backend/collector.py
@@ -17,7 +17,7 @@ from ray.tune.result import DEFAULT_RESULTS_DIR, JOB_META_FILE, \
     EXPR_PARAM_FILE, EXPR_RESULT_FILE, EXPR_META_FILE
 
 
-class CollectorService(object):
+class CollectorService:
     """Server implementation to monitor the log directory.
 
     The service will save the information of job and
diff --git a/python/ray/tune/checkpoint_manager.py b/python/ray/tune/checkpoint_manager.py
index 1ef0f6e20..c386d34d3 100644
--- a/python/ray/tune/checkpoint_manager.py
+++ b/python/ray/tune/checkpoint_manager.py
@@ -16,7 +16,7 @@ except NameError:
 logger = logging.getLogger(__name__)
 
 
-class Checkpoint(object):
+class Checkpoint:
     """Describes a checkpoint of trial state.
 
     Checkpoint may be saved in different storage.
@@ -54,7 +54,7 @@ class Checkpoint(object):
         return Checkpoint(Checkpoint.MEMORY, value)
 
 
-class QueueItem(object):
+class QueueItem:
     def __init__(self, priority, value):
         self.priority = priority
         self.value = value
@@ -69,7 +69,7 @@ class QueueItem(object):
         return self.priority < other.priority
 
 
-class CheckpointManager(object):
+class CheckpointManager:
     """Manages checkpoints on the driver for a trial."""
 
     def __init__(self, keep_checkpoints_num, checkpoint_score_attr):
diff --git a/python/ray/tune/experiment.py b/python/ray/tune/experiment.py
index cff078b43..422769018 100644
--- a/python/ray/tune/experiment.py
+++ b/python/ray/tune/experiment.py
@@ -34,7 +34,7 @@ def _raise_deprecation_note(deprecated, replacement, soft=False):
         raise DeprecationWarning(error_msg)
 
 
-class Experiment(object):
+class Experiment:
     """Tracks experiment specifications.
 
     Implicitly registers the Trainable if needed.
diff --git a/python/ray/tune/function_runner.py b/python/ray/tune/function_runner.py
index 6787ad6cf..ddbb85b95 100644
--- a/python/ray/tune/function_runner.py
+++ b/python/ray/tune/function_runner.py
@@ -24,7 +24,7 @@ ERROR_REPORT_TIMEOUT = 10
 ERROR_FETCH_TIMEOUT = 1
 
 
-class StatusReporter(object):
+class StatusReporter:
     """Object passed into your function that you can report status through.
 
     Example:
diff --git a/python/ray/tune/logger.py b/python/ray/tune/logger.py
index 5ee4ea3f5..1fd8dc0e1 100644
--- a/python/ray/tune/logger.py
+++ b/python/ray/tune/logger.py
@@ -26,7 +26,7 @@ tf = None
 VALID_SUMMARY_TYPES = [int, float, np.float32, np.float64, np.int32]
 
 
-class Logger(object):
+class Logger:
     """Logging interface for ray.tune.
 
     By default, the UnifiedLogger implementation is used which logs results in
diff --git a/python/ray/tune/progress_reporter.py b/python/ray/tune/progress_reporter.py
index 3d6d86b55..34a1de5fe 100644
--- a/python/ray/tune/progress_reporter.py
+++ b/python/ray/tune/progress_reporter.py
@@ -26,7 +26,7 @@ REPORTED_REPRESENTATIONS = {
 }
 
 
-class ProgressReporter(object):
+class ProgressReporter:
     # TODO(ujvl): Expose ProgressReporter in tune.run for custom reporting.
 
     def report(self, trial_runner):
diff --git a/python/ray/tune/ray_trial_executor.py b/python/ray/tune/ray_trial_executor.py
index 86cb6d5a7..1c763f37b 100644
--- a/python/ray/tune/ray_trial_executor.py
+++ b/python/ray/tune/ray_trial_executor.py
@@ -30,7 +30,7 @@ DEFAULT_GET_TIMEOUT = 30.0  # seconds
 TRIAL_START_ATTEMPTS = 3
 
 
-class _LocalWrapper(object):
+class _LocalWrapper:
     def __init__(self, result):
         self._result = result
 
diff --git a/python/ray/tune/registry.py b/python/ray/tune/registry.py
index 80d78e70f..9b5b178c9 100644
--- a/python/ray/tune/registry.py
+++ b/python/ray/tune/registry.py
@@ -98,7 +98,7 @@ def _make_key(category, key):
             key.encode("ascii"))
 
 
-class _Registry(object):
+class _Registry:
     def __init__(self):
         self._to_flush = {}
 
diff --git a/python/ray/tune/sample.py b/python/ray/tune/sample.py
index ae457f374..5cfb11c7f 100644
--- a/python/ray/tune/sample.py
+++ b/python/ray/tune/sample.py
@@ -8,7 +8,7 @@ import numpy as np
 logger = logging.getLogger(__name__)
 
 
-class sample_from(object):
+class sample_from:
     """Specify that tune should sample configuration values from this function.
 
     Arguments:
diff --git a/python/ray/tune/schedulers/hyperband.py b/python/ray/tune/schedulers/hyperband.py
index 2ed6ecfdf..66188028f 100644
--- a/python/ray/tune/schedulers/hyperband.py
+++ b/python/ray/tune/schedulers/hyperband.py
@@ -294,7 +294,7 @@ class HyperBandScheduler(FIFOScheduler):
         trial_runner.trial_executor.unpause_trial(trial)
 
 
-class Bracket(object):
+class Bracket:
     """Logical object for tracking Hyperband bracket progress. Keeps track
     of proper parameters as designated by HyperBand.
 
diff --git a/python/ray/tune/schedulers/pbt.py b/python/ray/tune/schedulers/pbt.py
index 67aab6b33..fa0f8c416 100644
--- a/python/ray/tune/schedulers/pbt.py
+++ b/python/ray/tune/schedulers/pbt.py
@@ -21,7 +21,7 @@ from ray.tune.trial import Trial, Checkpoint
 logger = logging.getLogger(__name__)
 
 
-class PBTTrialState(object):
+class PBTTrialState:
     """Internal PBT state tracked per-trial."""
 
     def __init__(self, trial):
diff --git a/python/ray/tune/schedulers/trial_scheduler.py b/python/ray/tune/schedulers/trial_scheduler.py
index 15fa3cb4c..2e85ac90f 100644
--- a/python/ray/tune/schedulers/trial_scheduler.py
+++ b/python/ray/tune/schedulers/trial_scheduler.py
@@ -5,7 +5,7 @@ from __future__ import print_function
 from ray.tune.trial import Trial
 
 
-class TrialScheduler(object):
+class TrialScheduler:
     """Interface for implementing a Trial Scheduler class."""
 
     CONTINUE = "CONTINUE"  #: Status for continuing trial execution
diff --git a/python/ray/tune/suggest/search.py b/python/ray/tune/suggest/search.py
index 89c0d26f9..f18f7e3ac 100644
--- a/python/ray/tune/suggest/search.py
+++ b/python/ray/tune/suggest/search.py
@@ -3,7 +3,7 @@ from __future__ import division
 from __future__ import print_function
 
 
-class SearchAlgorithm(object):
+class SearchAlgorithm:
     """Interface of an event handler API for hyperparameter search.
 
     Unlike TrialSchedulers, SearchAlgorithms will not have the ability
diff --git a/python/ray/tune/sync_client.py b/python/ray/tune/sync_client.py
index 9579af836..a6a1ef144 100644
--- a/python/ray/tune/sync_client.py
+++ b/python/ray/tune/sync_client.py
@@ -76,7 +76,7 @@ def get_cloud_sync_client(remote_path):
     return CommandBasedClient(template, template)
 
 
-class SyncClient(object):
+class SyncClient:
     """Client interface for interacting with remote storage options."""
 
     def sync_up(self, source, target):
diff --git a/python/ray/tune/syncer.py b/python/ray/tune/syncer.py
index 80d482b60..a4c69fc2b 100644
--- a/python/ray/tune/syncer.py
+++ b/python/ray/tune/syncer.py
@@ -60,7 +60,7 @@ def log_sync_template(options=""):
     return template.format(options=options, rsh=quote(rsh))
 
 
-class Syncer(object):
+class Syncer:
     def __init__(self, local_dir, remote_dir, sync_client=NOOP):
         """Syncs between two directories with the sync_function.
 
diff --git a/python/ray/tune/tests/test_api.py b/python/ray/tune/tests/test_api.py
index f7d34d662..8b0ed184e 100644
--- a/python/ray/tune/tests/test_api.py
+++ b/python/ray/tune/tests/test_api.py
@@ -192,7 +192,7 @@ class TrainableFunctionApiTest(unittest.TestCase):
         def train(config, reporter):
             pass
 
-        class A(object):
+        class A:
             pass
 
         class B(Trainable):
diff --git a/python/ray/tune/tests/test_commands.py b/python/ray/tune/tests/test_commands.py
index e82f21d34..6ce21dd77 100644
--- a/python/ray/tune/tests/test_commands.py
+++ b/python/ray/tune/tests/test_commands.py
@@ -20,7 +20,7 @@ from ray.tune import commands
 from ray.tune.result import CONFIG_PREFIX
 
 
-class Capturing(object):
+class Capturing:
     def __enter__(self):
         self._stdout = sys.stdout
         sys.stdout = self._stringio = StringIO()
diff --git a/python/ray/tune/tests/test_tune_restore.py b/python/ray/tune/tests/test_tune_restore.py
index b24a755a5..fb8218fc7 100644
--- a/python/ray/tune/tests/test_tune_restore.py
+++ b/python/ray/tune/tests/test_tune_restore.py
@@ -116,7 +116,7 @@ class AutoInitTest(unittest.TestCase):
         _register_all()
 
 
-class AbstractWarmStartTest(object):
+class AbstractWarmStartTest:
     def setUp(self):
         ray.init(local_mode=True)
         self.tmpdir = tempfile.mkdtemp()
diff --git a/python/ray/tune/track/session.py b/python/ray/tune/track/session.py
index 6f02c14cd..af9600729 100644
--- a/python/ray/tune/track/session.py
+++ b/python/ray/tune/track/session.py
@@ -18,7 +18,7 @@ class _ReporterHook(Logger):
         return self.tune_reporter(**metrics)
 
 
-class TrackSession(object):
+class TrackSession:
     """Manages results for a single session.
 
     Represents a single Trial in an experiment.
diff --git a/python/ray/tune/trainable.py b/python/ray/tune/trainable.py
index 98d529945..ae9c45df6 100644
--- a/python/ray/tune/trainable.py
+++ b/python/ray/tune/trainable.py
@@ -29,7 +29,7 @@ logger = logging.getLogger(__name__)
 SETUP_TIME_THRESHOLD = 10
 
 
-class Trainable(object):
+class Trainable:
     """Abstract class for trainable models, functions, etc.
 
     A call to ``train()`` on a trainable will execute one logical iteration of
diff --git a/python/ray/tune/trial.py b/python/ray/tune/trial.py
index 7ee547921..055352f0c 100644
--- a/python/ray/tune/trial.py
+++ b/python/ray/tune/trial.py
@@ -32,7 +32,7 @@ def date_str():
     return datetime.today().strftime("%Y-%m-%d_%H-%M-%S")
 
 
-class Location(object):
+class Location:
     """Describes the location at which Trial is placed to run."""
 
     def __init__(self, hostname=None, pid=None):
@@ -48,7 +48,7 @@ class Location(object):
             return "{}:{}".format(self.hostname, self.pid)
 
 
-class ExportFormat(object):
+class ExportFormat:
     """Describes the format to export the trial Trainable.
 
     This may correspond to different file formats based on the
@@ -73,7 +73,7 @@ class ExportFormat(object):
                                 export_formats[i])
 
 
-class Trial(object):
+class Trial:
     """A trial object holds the state for one model training run.
 
     Trials are themselves managed by the TrialRunner class, which implements
diff --git a/python/ray/tune/trial_executor.py b/python/ray/tune/trial_executor.py
index 859cb23eb..3230d5089 100644
--- a/python/ray/tune/trial_executor.py
+++ b/python/ray/tune/trial_executor.py
@@ -11,7 +11,7 @@ from ray.tune.error import TuneError
 logger = logging.getLogger(__name__)
 
 
-class TrialExecutor(object):
+class TrialExecutor:
     """Manages platform-specific details such as resource handling
     and starting/stopping trials.
     """
diff --git a/python/ray/tune/trial_runner.py b/python/ray/tune/trial_runner.py
index 80bbcbe57..f78bdc8ac 100644
--- a/python/ray/tune/trial_runner.py
+++ b/python/ray/tune/trial_runner.py
@@ -77,7 +77,7 @@ class _TuneFunctionDecoder(json.JSONDecoder):
         return cloudpickle.loads(hex_to_binary(obj["value"]))
 
 
-class TrialRunner(object):
+class TrialRunner:
     """A TrialRunner implements the event loop for scheduling trials on Ray.
 
     Example:
diff --git a/python/ray/tune/util.py b/python/ray/tune/util.py
index 99c1eace7..0a9dec3f6 100644
--- a/python/ray/tune/util.py
+++ b/python/ray/tune/util.py
@@ -111,7 +111,7 @@ def get_pinned_object(pinned_id):
     return ray.get(pinned_id)
 
 
-class warn_if_slow(object):
+class warn_if_slow:
     """Prints a warning if a given operation is slower than 100ms.
 
     Example:
diff --git a/python/ray/tune/web_server.py b/python/ray/tune/web_server.py
index b7e8783f2..9b9da1946 100644
--- a/python/ray/tune/web_server.py
+++ b/python/ray/tune/web_server.py
@@ -24,7 +24,7 @@ except ImportError:
                      "Be sure to install it on the client side.")
 
 
-class TuneClient(object):
+class TuneClient:
     """Client to interact with an ongoing Tune experiment.
 
     Requires a TuneServer to have started running.
diff --git a/python/ray/worker.py b/python/ray/worker.py
index 13a2593a7..0ab13218a 100644
--- a/python/ray/worker.py
+++ b/python/ray/worker.py
@@ -72,7 +72,7 @@ except ImportError:
     setproctitle = None
 
 
-class ActorCheckpointInfo(object):
+class ActorCheckpointInfo:
     """Information used to maintain actor checkpoints."""
 
     __slots__ = [
@@ -91,7 +91,7 @@ class ActorCheckpointInfo(object):
         self.checkpoint_ids = checkpoint_ids
 
 
-class Worker(object):
+class Worker:
     """A class used to define the control flow of a worker process.
 
     Note:
@@ -1707,7 +1707,7 @@ def remote(*args, **kwargs):
             return 1
 
         @ray.remote
-        class Foo(object):
+        class Foo:
             def method(self):
                 return 1
 
@@ -1743,7 +1743,7 @@ def remote(*args, **kwargs):
             return 1, 2
 
         @ray.remote(num_cpus=2, resources={"CustomResource": 1})
-        class Foo(object):
+        class Foo:
             def method(self):
                 return 1
 
@@ -1759,7 +1759,7 @@ def remote(*args, **kwargs):
         g = f.options(num_gpus=2, max_calls=None)
 
         @ray.remote(num_cpus=2, resources={"CustomResource": 1})
-        class Foo(object):
+        class Foo:
             def method(self):
                 return 1
         Bar = Foo.options(num_cpus=1, resources=None)
diff --git a/rllib/agents/a3c/a3c_tf_policy.py b/rllib/agents/a3c/a3c_tf_policy.py
index 2bb5d8bae..6aed17c0a 100644
--- a/rllib/agents/a3c/a3c_tf_policy.py
+++ b/rllib/agents/a3c/a3c_tf_policy.py
@@ -17,7 +17,7 @@ from ray.rllib.utils import try_import_tf
 tf = try_import_tf()
 
 
-class A3CLoss(object):
+class A3CLoss:
     def __init__(self,
                  action_dist,
                  actions,
@@ -73,7 +73,7 @@ def add_value_function_fetch(policy):
     return {SampleBatch.VF_PREDS: policy.model.value_function()}
 
 
-class ValueNetworkMixin(object):
+class ValueNetworkMixin:
     def __init__(self):
         @make_tf_callable(self.get_session())
         def value(ob, prev_action, prev_reward, *state):
@@ -94,8 +94,7 @@ def stats(policy, train_batch):
         "cur_lr": tf.cast(policy.cur_lr, tf.float64),
         "policy_loss": policy.loss.pi_loss,
         "policy_entropy": policy.loss.entropy,
-        "var_gnorm": tf.global_norm(
-            list(policy.model.trainable_variables())),
+        "var_gnorm": tf.global_norm(list(policy.model.trainable_variables())),
         "vf_loss": policy.loss.vf_loss,
     }
 
diff --git a/rllib/agents/a3c/a3c_torch_policy.py b/rllib/agents/a3c/a3c_torch_policy.py
index f12e817e0..e69b13a14 100644
--- a/rllib/agents/a3c/a3c_torch_policy.py
+++ b/rllib/agents/a3c/a3c_torch_policy.py
@@ -69,7 +69,7 @@ def torch_optimizer(policy, config):
     return torch.optim.Adam(policy.model.parameters(), lr=config["lr"])
 
 
-class ValueNetworkMixin(object):
+class ValueNetworkMixin:
     def _value(self, obs):
         obs = torch.from_numpy(obs).float().unsqueeze(0).to(self.device)
         _ = self.model({"obs": obs}, [], [1])
diff --git a/rllib/agents/ars/ars.py b/rllib/agents/ars/ars.py
index 4330f0d90..f461e24c5 100644
--- a/rllib/agents/ars/ars.py
+++ b/rllib/agents/ars/ars.py
@@ -55,7 +55,7 @@ def create_shared_noise(count):
     return noise
 
 
-class SharedNoiseTable(object):
+class SharedNoiseTable:
     def __init__(self, noise):
         self.noise = noise
         assert self.noise.dtype == np.float32
@@ -72,7 +72,7 @@ class SharedNoiseTable(object):
 
 
 @ray.remote
-class Worker(object):
+class Worker:
     def __init__(self, config, env_creator, noise, min_task_runtime=0.2):
         self.min_task_runtime = min_task_runtime
         self.config = config
diff --git a/rllib/agents/ars/optimizers.py b/rllib/agents/ars/optimizers.py
index 0e6420bb2..5297c1ad7 100644
--- a/rllib/agents/ars/optimizers.py
+++ b/rllib/agents/ars/optimizers.py
@@ -8,7 +8,7 @@ from __future__ import print_function
 import numpy as np
 
 
-class Optimizer(object):
+class Optimizer:
     def __init__(self, policy):
         self.policy = policy
         self.dim = policy.num_params
diff --git a/rllib/agents/ars/policies.py b/rllib/agents/ars/policies.py
index ce3e837a3..1ce059410 100644
--- a/rllib/agents/ars/policies.py
+++ b/rllib/agents/ars/policies.py
@@ -56,7 +56,7 @@ def rollout(policy, env, timestep_limit=None, add_noise=False, offset=0):
     return rews, t
 
 
-class GenericPolicy(object):
+class GenericPolicy:
     def __init__(self,
                  sess,
                  action_space,
diff --git a/rllib/agents/ddpg/ddpg_policy.py b/rllib/agents/ddpg/ddpg_policy.py
index 95e4bd121..caf2754ef 100644
--- a/rllib/agents/ddpg/ddpg_policy.py
+++ b/rllib/agents/ddpg/ddpg_policy.py
@@ -32,7 +32,7 @@ TWIN_Q_TARGET_SCOPE = "twin_target_critic"
 PRIO_WEIGHTS = "weights"
 
 
-class DDPGPostprocessing(object):
+class DDPGPostprocessing:
     """Implements n-step learning and param noise adjustments."""
 
     @override(Policy)
diff --git a/rllib/agents/dqn/dqn_policy.py b/rllib/agents/dqn/dqn_policy.py
index dbfb6488f..044eb9cd2 100644
--- a/rllib/agents/dqn/dqn_policy.py
+++ b/rllib/agents/dqn/dqn_policy.py
@@ -30,7 +30,7 @@ Q_TARGET_SCOPE = "target_q_func"
 PRIO_WEIGHTS = "weights"
 
 
-class QLoss(object):
+class QLoss:
     def __init__(self,
                  q_t_selected,
                  q_logits_t_selected,
@@ -107,7 +107,7 @@ class QLoss(object):
             }
 
 
-class QValuePolicy(object):
+class QValuePolicy:
     def __init__(self, q_values, observations, num_actions, cur_epsilon,
                  softmax, softmax_temp, model_config):
         if softmax:
@@ -135,7 +135,7 @@ class QValuePolicy(object):
         self.action_prob = None
 
 
-class ComputeTDErrorMixin(object):
+class ComputeTDErrorMixin:
     def __init__(self):
         @make_tf_callable(self.get_session(), dynamic_shape=True)
         def compute_td_error(obs_t, act_t, rew_t, obs_tp1, done_mask,
diff --git a/rllib/agents/dqn/simple_q_policy.py b/rllib/agents/dqn/simple_q_policy.py
index 996a62786..ee8dfe24c 100644
--- a/rllib/agents/dqn/simple_q_policy.py
+++ b/rllib/agents/dqn/simple_q_policy.py
@@ -25,7 +25,7 @@ Q_SCOPE = "q_func"
 Q_TARGET_SCOPE = "target_q_func"
 
 
-class ExplorationStateMixin(object):
+class ExplorationStateMixin:
     def __init__(self, obs_space, action_space, config):
         # Python value, should always be same as the TF variable
         self.cur_epsilon_value = 1.0
@@ -55,7 +55,7 @@ class ExplorationStateMixin(object):
         self.set_epsilon(state[1])
 
 
-class TargetNetworkMixin(object):
+class TargetNetworkMixin:
     def __init__(self, obs_space, action_space, config):
         @make_tf_callable(self.get_session())
         def do_update():
diff --git a/rllib/agents/es/es.py b/rllib/agents/es/es.py
index f5338a632..e4c8ddc2b 100644
--- a/rllib/agents/es/es.py
+++ b/rllib/agents/es/es.py
@@ -55,7 +55,7 @@ def create_shared_noise(count):
     return noise
 
 
-class SharedNoiseTable(object):
+class SharedNoiseTable:
     def __init__(self, noise):
         self.noise = noise
         assert self.noise.dtype == np.float32
@@ -68,7 +68,7 @@ class SharedNoiseTable(object):
 
 
 @ray.remote
-class Worker(object):
+class Worker:
     def __init__(self,
                  config,
                  policy_params,
diff --git a/rllib/agents/es/optimizers.py b/rllib/agents/es/optimizers.py
index 3b48f7393..726a90635 100644
--- a/rllib/agents/es/optimizers.py
+++ b/rllib/agents/es/optimizers.py
@@ -8,7 +8,7 @@ from __future__ import print_function
 import numpy as np
 
 
-class Optimizer(object):
+class Optimizer:
     def __init__(self, pi):
         self.pi = pi
         self.dim = pi.num_params
diff --git a/rllib/agents/es/policies.py b/rllib/agents/es/policies.py
index 3ddb4dbed..8184bab5b 100644
--- a/rllib/agents/es/policies.py
+++ b/rllib/agents/es/policies.py
@@ -41,7 +41,7 @@ def rollout(policy, env, timestep_limit=None, add_noise=False):
     return rews, t
 
 
-class GenericPolicy(object):
+class GenericPolicy:
     def __init__(self, sess, action_space, obs_space, preprocessor,
                  observation_filter, model_options, action_noise_std):
         self.sess = sess
diff --git a/rllib/agents/impala/impala.py b/rllib/agents/impala/impala.py
index e83c4ca65..6b3ef79e1 100644
--- a/rllib/agents/impala/impala.py
+++ b/rllib/agents/impala/impala.py
@@ -139,7 +139,7 @@ def make_aggregators_and_optimizer(workers, config):
     return optimizer
 
 
-class OverrideDefaultResourceRequest(object):
+class OverrideDefaultResourceRequest:
     @classmethod
     @override(Trainable)
     def default_resource_request(cls, config):
diff --git a/rllib/agents/impala/vtrace_policy.py b/rllib/agents/impala/vtrace_policy.py
index 7e51a9959..17212aff4 100644
--- a/rllib/agents/impala/vtrace_policy.py
+++ b/rllib/agents/impala/vtrace_policy.py
@@ -27,7 +27,7 @@ logger = logging.getLogger(__name__)
 BEHAVIOUR_LOGITS = "behaviour_logits"
 
 
-class VTraceLoss(object):
+class VTraceLoss:
     def __init__(self,
                  actions,
                  actions_logp,
diff --git a/rllib/agents/marwil/marwil_policy.py b/rllib/agents/marwil/marwil_policy.py
index 72b8a2393..f271b78d2 100644
--- a/rllib/agents/marwil/marwil_policy.py
+++ b/rllib/agents/marwil/marwil_policy.py
@@ -21,13 +21,13 @@ POLICY_SCOPE = "p_func"
 VALUE_SCOPE = "v_func"
 
 
-class ValueLoss(object):
+class ValueLoss:
     def __init__(self, state_values, cumulative_rewards):
         self.loss = 0.5 * tf.reduce_mean(
             tf.square(state_values - cumulative_rewards))
 
 
-class ReweightedImitationLoss(object):
+class ReweightedImitationLoss:
     def __init__(self, state_values, cumulative_rewards, logits, actions,
                  action_space, beta, model):
         ma_adv_norm = tf.get_variable(
@@ -56,7 +56,7 @@ class ReweightedImitationLoss(object):
             tf.stop_gradient(exp_advs) * logprobs)
 
 
-class MARWILPostprocessing(object):
+class MARWILPostprocessing:
     """Adds the advantages field to the trajectory."""
 
     @override(Policy)
diff --git a/rllib/agents/ppo/appo_policy.py b/rllib/agents/ppo/appo_policy.py
index ce439a99e..1d2f899e5 100644
--- a/rllib/agents/ppo/appo_policy.py
+++ b/rllib/agents/ppo/appo_policy.py
@@ -34,7 +34,7 @@ TARGET_POLICY_SCOPE = "target_func"
 logger = logging.getLogger(__name__)
 
 
-class PPOSurrogateLoss(object):
+class PPOSurrogateLoss:
     """Loss used when V-trace is disabled.
 
     Arguments:
@@ -97,7 +97,7 @@ class PPOSurrogateLoss(object):
             self.total_loss += cur_kl_coeff * self.mean_kl
 
 
-class VTraceSurrogateLoss(object):
+class VTraceSurrogateLoss:
     def __init__(self,
                  actions,
                  prev_actions_logp,
@@ -407,7 +407,7 @@ def add_values_and_logits(policy):
     return out
 
 
-class TargetNetworkMixin(object):
+class TargetNetworkMixin:
     def __init__(self, obs_space, action_space, config):
         """Target Network is updated by the master learner every
         trainer.update_target_frequency steps. All worker batches
diff --git a/rllib/agents/ppo/ppo_policy.py b/rllib/agents/ppo/ppo_policy.py
index 16bb7dbd0..4821c97db 100644
--- a/rllib/agents/ppo/ppo_policy.py
+++ b/rllib/agents/ppo/ppo_policy.py
@@ -23,7 +23,7 @@ logger = logging.getLogger(__name__)
 BEHAVIOUR_LOGITS = "behaviour_logits"
 
 
-class PPOLoss(object):
+class PPOLoss:
     def __init__(self,
                  action_space,
                  dist_class,
@@ -211,7 +211,7 @@ def clip_gradients(policy, optimizer, loss):
         return optimizer.compute_gradients(loss, variables)
 
 
-class KLCoeffMixin(object):
+class KLCoeffMixin:
     def __init__(self, config):
         # KL Coefficient
         self.kl_coeff_val = config["kl_coeff"]
@@ -232,7 +232,7 @@ class KLCoeffMixin(object):
         return self.kl_coeff_val
 
 
-class ValueNetworkMixin(object):
+class ValueNetworkMixin:
     def __init__(self, obs_space, action_space, config):
         if config["use_gae"]:
 
diff --git a/rllib/agents/sac/sac_policy.py b/rllib/agents/sac/sac_policy.py
index da5cc14d1..99357b7a7 100644
--- a/rllib/agents/sac/sac_policy.py
+++ b/rllib/agents/sac/sac_policy.py
@@ -320,7 +320,7 @@ def stats(policy, train_batch):
     }
 
 
-class ExplorationStateMixin(object):
+class ExplorationStateMixin:
     def __init__(self, obs_space, action_space, config):
         self.stochastic = tf.get_variable(
             initializer=tf.constant_initializer(config["exploration_enabled"]),
@@ -333,7 +333,7 @@ class ExplorationStateMixin(object):
         pass
 
 
-class ActorCriticOptimizerMixin(object):
+class ActorCriticOptimizerMixin:
     def __init__(self, config):
         # create global step for counting the number of update operations
         self.global_step = tf.train.get_or_create_global_step()
@@ -353,7 +353,7 @@ class ActorCriticOptimizerMixin(object):
             learning_rate=config["optimization"]["entropy_learning_rate"])
 
 
-class ComputeTDErrorMixin(object):
+class ComputeTDErrorMixin:
     def __init__(self):
         @make_tf_callable(self.get_session(), dynamic_shape=True)
         def compute_td_error(obs_t, act_t, rew_t, obs_tp1, done_mask,
@@ -374,7 +374,7 @@ class ComputeTDErrorMixin(object):
         self.compute_td_error = compute_td_error
 
 
-class TargetNetworkMixin(object):
+class TargetNetworkMixin:
     def __init__(self, config):
         @make_tf_callable(self.get_session())
         def update_target_fn(tau):
diff --git a/rllib/contrib/alpha_zero/core/mcts.py b/rllib/contrib/alpha_zero/core/mcts.py
index c31538236..fd5ce71a0 100644
--- a/rllib/contrib/alpha_zero/core/mcts.py
+++ b/rllib/contrib/alpha_zero/core/mcts.py
@@ -100,7 +100,7 @@ class Node:
             current = current.parent
 
 
-class RootParentNode(object):
+class RootParentNode:
     def __init__(self, env):
         self.parent = None
         self.child_total_value = collections.defaultdict(float)
diff --git a/rllib/contrib/maddpg/maddpg_policy.py b/rllib/contrib/maddpg/maddpg_policy.py
index d0acdaa7c..60befde05 100644
--- a/rllib/contrib/maddpg/maddpg_policy.py
+++ b/rllib/contrib/maddpg/maddpg_policy.py
@@ -23,7 +23,7 @@ tf = try_import_tf()
 tfp = try_import_tfp()
 
 
-class MADDPGPostprocessing(object):
+class MADDPGPostprocessing:
     """Implements agentwise termination signal and n-step learning."""
 
     @override(Policy)
diff --git a/rllib/env/base_env.py b/rllib/env/base_env.py
index a36c3e228..f862e5c2a 100644
--- a/rllib/env/base_env.py
+++ b/rllib/env/base_env.py
@@ -12,7 +12,7 @@ ASYNC_RESET_RETURN = "async_reset_return"
 
 
 @PublicAPI
-class BaseEnv(object):
+class BaseEnv:
     """The lowest-level env interface used by RLlib for sampling.
 
     BaseEnv models multiple agents executing asynchronously in multiple
@@ -412,7 +412,7 @@ class _MultiAgentEnvToBaseEnv(BaseEnv):
         return [state.env for state in self.env_states]
 
 
-class _MultiAgentEnvState(object):
+class _MultiAgentEnvState:
     def __init__(self, env):
         assert isinstance(env, MultiAgentEnv)
         self.env = env
diff --git a/rllib/env/external_env.py b/rllib/env/external_env.py
index 46d15a2de..562b87418 100644
--- a/rllib/env/external_env.py
+++ b/rllib/env/external_env.py
@@ -181,7 +181,7 @@ class ExternalEnv(threading.Thread):
         return self._episodes[episode_id]
 
 
-class _ExternalEnvEpisode(object):
+class _ExternalEnvEpisode:
     """Tracked state for each active episode."""
 
     def __init__(self,
diff --git a/rllib/env/multi_agent_env.py b/rllib/env/multi_agent_env.py
index d770215be..7b1f5ce5d 100644
--- a/rllib/env/multi_agent_env.py
+++ b/rllib/env/multi_agent_env.py
@@ -6,7 +6,7 @@ from ray.rllib.utils.annotations import PublicAPI
 
 
 @PublicAPI
-class MultiAgentEnv(object):
+class MultiAgentEnv:
     """An environment that hosts multiple independent agents.
 
     Agents are identified by (string) agent ids. Note that these "agents" here
diff --git a/rllib/env/remote_vector_env.py b/rllib/env/remote_vector_env.py
index 7c891cf3c..d6d6da154 100644
--- a/rllib/env/remote_vector_env.py
+++ b/rllib/env/remote_vector_env.py
@@ -89,7 +89,7 @@ class RemoteVectorEnv(BaseEnv):
 
 
 @ray.remote(num_cpus=0)
-class _RemoteMultiAgentEnv(object):
+class _RemoteMultiAgentEnv:
     """Wrapper class for making a multi-agent env a remote actor."""
 
     def __init__(self, make_env, i):
@@ -108,7 +108,7 @@ class _RemoteMultiAgentEnv(object):
 
 
 @ray.remote(num_cpus=0)
-class _RemoteSingleAgentEnv(object):
+class _RemoteSingleAgentEnv:
     """Wrapper class for making a gym env a remote actor."""
 
     def __init__(self, make_env, i):
diff --git a/rllib/env/vector_env.py b/rllib/env/vector_env.py
index d0df24177..66bd8e737 100644
--- a/rllib/env/vector_env.py
+++ b/rllib/env/vector_env.py
@@ -11,7 +11,7 @@ logger = logging.getLogger(__name__)
 
 
 @PublicAPI
-class VectorEnv(object):
+class VectorEnv:
     """An environment that supports batch evaluation.
 
     Subclasses must define the following attributes:
diff --git a/rllib/evaluation/episode.py b/rllib/evaluation/episode.py
index 2ee1d58ee..ab4746b79 100644
--- a/rllib/evaluation/episode.py
+++ b/rllib/evaluation/episode.py
@@ -12,7 +12,7 @@ from ray.rllib.utils.annotations import DeveloperAPI
 
 
 @DeveloperAPI
-class MultiAgentEpisode(object):
+class MultiAgentEpisode:
     """Tracks the current state of a (possibly multi-agent) episode.
 
     Attributes:
diff --git a/rllib/evaluation/interface.py b/rllib/evaluation/interface.py
index 06fa9f94e..739aa55b1 100644
--- a/rllib/evaluation/interface.py
+++ b/rllib/evaluation/interface.py
@@ -8,7 +8,7 @@ from ray.rllib.utils.annotations import DeveloperAPI
 
 
 @DeveloperAPI
-class EvaluatorInterface(object):
+class EvaluatorInterface:
     """This is the interface between policy optimizers and policy evaluation.
 
     See also: RolloutWorker
diff --git a/rllib/evaluation/postprocessing.py b/rllib/evaluation/postprocessing.py
index 6ca49f228..46a19b929 100644
--- a/rllib/evaluation/postprocessing.py
+++ b/rllib/evaluation/postprocessing.py
@@ -12,7 +12,7 @@ def discount(x, gamma):
     return scipy.signal.lfilter([1], [1, -gamma], x[::-1], axis=0)[::-1]
 
 
-class Postprocessing(object):
+class Postprocessing:
     """Constant definitions for postprocessing."""
 
     ADVANTAGES = "advantages"
diff --git a/rllib/evaluation/sample_batch_builder.py b/rllib/evaluation/sample_batch_builder.py
index 987fb521d..85e608c31 100644
--- a/rllib/evaluation/sample_batch_builder.py
+++ b/rllib/evaluation/sample_batch_builder.py
@@ -21,7 +21,7 @@ def to_float_array(v):
 
 
 @PublicAPI
-class SampleBatchBuilder(object):
+class SampleBatchBuilder:
     """Util to build a SampleBatch incrementally.
 
     For efficiency, SampleBatches hold values in column form (as arrays).
@@ -66,7 +66,7 @@ class SampleBatchBuilder(object):
 
 
 @DeveloperAPI
-class MultiAgentSampleBatchBuilder(object):
+class MultiAgentSampleBatchBuilder:
     """Util to build SampleBatches for each policy in a multi-agent env.
 
     Input data is per-agent, while output data is per-policy. There is an M:N
diff --git a/rllib/evaluation/sampler.py b/rllib/evaluation/sampler.py
index 41a1b4389..5917397ad 100644
--- a/rllib/evaluation/sampler.py
+++ b/rllib/evaluation/sampler.py
@@ -31,7 +31,7 @@ PolicyEvalData = namedtuple("PolicyEvalData", [
 ])
 
 
-class PerfStats(object):
+class PerfStats:
     """Sampler perf stats that will be included in rollout metrics."""
 
     def __init__(self):
diff --git a/rllib/evaluation/worker_set.py b/rllib/evaluation/worker_set.py
index 7a23caa08..e3c39c48e 100644
--- a/rllib/evaluation/worker_set.py
+++ b/rllib/evaluation/worker_set.py
@@ -19,7 +19,7 @@ logger = logging.getLogger(__name__)
 
 
 @DeveloperAPI
-class WorkerSet(object):
+class WorkerSet:
     """Represents a set of RolloutWorkers.
 
     There must be one local worker copy, and zero or more remote workers.
diff --git a/rllib/examples/centralized_critic.py b/rllib/examples/centralized_critic.py
index 90f704046..a9aa5fc23 100644
--- a/rllib/examples/centralized_critic.py
+++ b/rllib/examples/centralized_critic.py
@@ -84,7 +84,7 @@ class CentralizedCriticModel(TFModelV2):
         return self.model.value_function()  # not used
 
 
-class CentralizedValueMixin(object):
+class CentralizedValueMixin:
     """Add method to evaluate the central value function from the model."""
 
     def __init__(self):
diff --git a/rllib/models/action_dist.py b/rllib/models/action_dist.py
index f5a5f1e3c..0b88547d2 100644
--- a/rllib/models/action_dist.py
+++ b/rllib/models/action_dist.py
@@ -6,7 +6,7 @@ from ray.rllib.utils.annotations import DeveloperAPI
 
 
 @DeveloperAPI
-class ActionDistribution(object):
+class ActionDistribution:
     """The policy action distribution of an agent.
 
     Attributes:
diff --git a/rllib/models/catalog.py b/rllib/models/catalog.py
index a43c7e372..c8f1711ff 100644
--- a/rllib/models/catalog.py
+++ b/rllib/models/catalog.py
@@ -92,7 +92,7 @@ MODEL_DEFAULTS = {
 
 
 @PublicAPI
-class ModelCatalog(object):
+class ModelCatalog:
     """Registry of models, preprocessors, and action distributions for envs.
 
     Examples:
diff --git a/rllib/models/model.py b/rllib/models/model.py
index 996179630..bfc2e7a65 100644
--- a/rllib/models/model.py
+++ b/rllib/models/model.py
@@ -17,7 +17,7 @@ torch, _ = try_import_torch()
 logger = logging.getLogger(__name__)
 
 
-class Model(object):
+class Model:
     """This class is deprecated, please use TFModelV2 instead."""
 
     def __init__(self,
diff --git a/rllib/models/modelv2.py b/rllib/models/modelv2.py
index 568963da0..14165607f 100644
--- a/rllib/models/modelv2.py
+++ b/rllib/models/modelv2.py
@@ -8,7 +8,7 @@ from ray.rllib.utils.annotations import PublicAPI
 
 
 @PublicAPI
-class ModelV2(object):
+class ModelV2:
     """Defines a Keras-style abstract network model for use with RLlib.
 
     Custom models should extend either TFModelV2 or TorchModelV2 instead of
@@ -205,7 +205,7 @@ class ModelV2(object):
         return NullContextManager()
 
 
-class NullContextManager(object):
+class NullContextManager:
     """No-op context manager"""
 
     def __init__(self):
diff --git a/rllib/models/preprocessors.py b/rllib/models/preprocessors.py
index e0a33d9b7..2bf567963 100644
--- a/rllib/models/preprocessors.py
+++ b/rllib/models/preprocessors.py
@@ -18,7 +18,7 @@ logger = logging.getLogger(__name__)
 
 
 @PublicAPI
-class Preprocessor(object):
+class Preprocessor:
     """Defines an abstract observation preprocessor function.
 
     Attributes:
diff --git a/rllib/offline/input_reader.py b/rllib/offline/input_reader.py
index 163a15e1b..240c4b3b1 100644
--- a/rllib/offline/input_reader.py
+++ b/rllib/offline/input_reader.py
@@ -16,7 +16,7 @@ logger = logging.getLogger(__name__)
 
 
 @PublicAPI
-class InputReader(object):
+class InputReader:
     """Input object for loading experiences in policy evaluation."""
 
     @PublicAPI
diff --git a/rllib/offline/io_context.py b/rllib/offline/io_context.py
index 58f7f03c5..183bd5994 100644
--- a/rllib/offline/io_context.py
+++ b/rllib/offline/io_context.py
@@ -8,7 +8,7 @@ from ray.rllib.utils.annotations import PublicAPI
 
 
 @PublicAPI
-class IOContext(object):
+class IOContext:
     """Attributes to pass to input / output class constructors.
 
     RLlib auto-sets these attributes when constructing input / output classes.
diff --git a/rllib/offline/off_policy_estimator.py b/rllib/offline/off_policy_estimator.py
index 9d369f715..c90bcaec9 100644
--- a/rllib/offline/off_policy_estimator.py
+++ b/rllib/offline/off_policy_estimator.py
@@ -15,7 +15,7 @@ OffPolicyEstimate = namedtuple("OffPolicyEstimate",
 
 
 @DeveloperAPI
-class OffPolicyEstimator(object):
+class OffPolicyEstimator:
     """Interface for an off policy reward estimator."""
 
     @DeveloperAPI
diff --git a/rllib/offline/output_writer.py b/rllib/offline/output_writer.py
index 003f073ec..13033b240 100644
--- a/rllib/offline/output_writer.py
+++ b/rllib/offline/output_writer.py
@@ -7,7 +7,7 @@ from ray.rllib.utils.annotations import PublicAPI
 
 
 @PublicAPI
-class OutputWriter(object):
+class OutputWriter:
     """Writer object for saving experiences from policy evaluation."""
 
     @PublicAPI
diff --git a/rllib/optimizers/aso_aggregator.py b/rllib/optimizers/aso_aggregator.py
index fc6d64071..7c5452bf0 100644
--- a/rllib/optimizers/aso_aggregator.py
+++ b/rllib/optimizers/aso_aggregator.py
@@ -13,7 +13,7 @@ from ray.rllib.utils.annotations import override
 from ray.rllib.utils.memory import ray_get_and_free
 
 
-class Aggregator(object):
+class Aggregator:
     """An aggregator collects and processes samples from workers.
 
     This class is used to abstract away the strategy for sample collection.
@@ -52,7 +52,7 @@ class Aggregator(object):
         raise NotImplementedError
 
 
-class AggregationWorkerBase(object):
+class AggregationWorkerBase:
     """Aggregators should extend from this class."""
 
     def __init__(self, initial_weights_obj_id, remote_workers,
diff --git a/rllib/optimizers/aso_minibatch_buffer.py b/rllib/optimizers/aso_minibatch_buffer.py
index b196f8aca..d9fe447de 100644
--- a/rllib/optimizers/aso_minibatch_buffer.py
+++ b/rllib/optimizers/aso_minibatch_buffer.py
@@ -5,7 +5,7 @@ from __future__ import division
 from __future__ import print_function
 
 
-class MinibatchBuffer(object):
+class MinibatchBuffer:
     """Ring buffer of recent data batches for minibatch SGD.
 
     This is for use with AsyncSamplesOptimizer.
diff --git a/rllib/optimizers/async_replay_optimizer.py b/rllib/optimizers/async_replay_optimizer.py
index 199f66086..1ba650de5 100644
--- a/rllib/optimizers/async_replay_optimizer.py
+++ b/rllib/optimizers/async_replay_optimizer.py
@@ -256,7 +256,7 @@ class AsyncReplayOptimizer(PolicyOptimizer):
 
 
 @ray.remote(num_cpus=0)
-class ReplayActor(object):
+class ReplayActor:
     """A replay buffer shard.
 
     Ray actors are single-threaded, so for scalability multiple replay actors
@@ -344,7 +344,7 @@ class ReplayActor(object):
 # note: we set num_cpus=0 to avoid failing to create replay actors when
 # resources are fragmented. This isn't ideal.
 @ray.remote(num_cpus=0)
-class BatchReplayActor(object):
+class BatchReplayActor:
     """The batch replay version of the replay actor.
 
     This allows for RNN models, but ignores prioritization params.
diff --git a/rllib/optimizers/multi_gpu_impl.py b/rllib/optimizers/multi_gpu_impl.py
index aa7340ab8..3d27d6dbc 100644
--- a/rllib/optimizers/multi_gpu_impl.py
+++ b/rllib/optimizers/multi_gpu_impl.py
@@ -16,7 +16,7 @@ TOWER_SCOPE_NAME = "tower"
 logger = logging.getLogger(__name__)
 
 
-class LocalSyncParallelOptimizer(object):
+class LocalSyncParallelOptimizer:
     """Optimizer that runs in parallel across multiple local devices.
 
     LocalSyncParallelOptimizer automatically splits up and loads training data
diff --git a/rllib/optimizers/policy_optimizer.py b/rllib/optimizers/policy_optimizer.py
index 3d0190dcc..d46f6e9a6 100644
--- a/rllib/optimizers/policy_optimizer.py
+++ b/rllib/optimizers/policy_optimizer.py
@@ -11,7 +11,7 @@ logger = logging.getLogger(__name__)
 
 
 @DeveloperAPI
-class PolicyOptimizer(object):
+class PolicyOptimizer:
     """Policy optimizers encapsulate distributed RL optimization strategies.
 
     Policy optimizers serve as the "control plane" of algorithms.
diff --git a/rllib/optimizers/replay_buffer.py b/rllib/optimizers/replay_buffer.py
index 90e1f87e3..0b7f0979b 100644
--- a/rllib/optimizers/replay_buffer.py
+++ b/rllib/optimizers/replay_buffer.py
@@ -13,7 +13,7 @@ from ray.rllib.utils.window_stat import WindowStat
 
 
 @DeveloperAPI
-class ReplayBuffer(object):
+class ReplayBuffer:
     @DeveloperAPI
     def __init__(self, size):
         """Create Prioritized Replay buffer.
diff --git a/rllib/optimizers/segment_tree.py b/rllib/optimizers/segment_tree.py
index 8f0a0b5a5..271965bac 100644
--- a/rllib/optimizers/segment_tree.py
+++ b/rllib/optimizers/segment_tree.py
@@ -5,7 +5,7 @@ from __future__ import print_function
 import operator
 
 
-class SegmentTree(object):
+class SegmentTree:
     def __init__(self, capacity, operation, neutral_element):
         """Build a Segment Tree data structure.
 
diff --git a/rllib/policy/eager_tf_policy.py b/rllib/policy/eager_tf_policy.py
index 3cb58ec46..461dde7e8 100644
--- a/rllib/policy/eager_tf_policy.py
+++ b/rllib/policy/eager_tf_policy.py
@@ -409,7 +409,7 @@ def build_eager_tf_policy(name,
 
             if gradients_fn:
 
-                class OptimizerWrapper(object):
+                class OptimizerWrapper:
                     def __init__(self, tape):
                         self.tape = tape
 
diff --git a/rllib/policy/policy.py b/rllib/policy/policy.py
index 2458ce5b0..0593d5c6a 100644
--- a/rllib/policy/policy.py
+++ b/rllib/policy/policy.py
@@ -24,7 +24,7 @@ class TupleActions(namedtuple("TupleActions", ["batches"])):
 
 
 @DeveloperAPI
-class Policy(object):
+class Policy:
     """An agent policy and loss, i.e., a TFPolicy or other subclass.
 
     This object defines how to act in the environment, and also losses used to
diff --git a/rllib/policy/sample_batch.py b/rllib/policy/sample_batch.py
index fbd4b7e37..8e177e42d 100644
--- a/rllib/policy/sample_batch.py
+++ b/rllib/policy/sample_batch.py
@@ -15,7 +15,7 @@ DEFAULT_POLICY_ID = "default_policy"
 
 
 @PublicAPI
-class SampleBatch(object):
+class SampleBatch:
     """Wrapper around a dictionary with string keys and array-like values.
 
     For example, {"obs": [1, 2, 3], "reward": [0, -1, 1]} is a batch of three
@@ -231,7 +231,7 @@ class SampleBatch(object):
 
 
 @PublicAPI
-class MultiAgentBatch(object):
+class MultiAgentBatch:
     """A batch of experiences from multiple policies in the environment.
 
     Attributes:
diff --git a/rllib/policy/tf_policy.py b/rllib/policy/tf_policy.py
index 67d51a42b..e8c654e4c 100644
--- a/rllib/policy/tf_policy.py
+++ b/rllib/policy/tf_policy.py
@@ -564,7 +564,7 @@ class TFPolicy(Policy):
 
 
 @DeveloperAPI
-class LearningRateSchedule(object):
+class LearningRateSchedule:
     """Mixin for TFPolicy that adds a learning rate schedule."""
 
     @DeveloperAPI
@@ -589,7 +589,7 @@ class LearningRateSchedule(object):
 
 
 @DeveloperAPI
-class EntropyCoeffSchedule(object):
+class EntropyCoeffSchedule:
     """Mixin for TFPolicy that adds entropy coeff decay."""
 
     @DeveloperAPI
diff --git a/rllib/tests/mock_worker.py b/rllib/tests/mock_worker.py
index b6b2e9773..9f34e3b7e 100644
--- a/rllib/tests/mock_worker.py
+++ b/rllib/tests/mock_worker.py
@@ -8,7 +8,7 @@ from ray.rllib.evaluation import SampleBatch
 from ray.rllib.utils.filter import MeanStdFilter
 
 
-class _MockWorker(object):
+class _MockWorker:
     def __init__(self, sample_count=10):
         self._weights = np.array([-10, -10, -10, -10])
         self._grad = np.array([1, 1, 1, 1])
diff --git a/rllib/tests/test_catalog.py b/rllib/tests/test_catalog.py
index 6fc7505ab..b8287ad65 100644
--- a/rllib/tests/test_catalog.py
+++ b/rllib/tests/test_catalog.py
@@ -65,7 +65,7 @@ class ModelCatalogTest(unittest.TestCase):
     def testTuplePreprocessor(self):
         ray.init(object_store_memory=1000 * 1024 * 1024)
 
-        class TupleEnv(object):
+        class TupleEnv:
             def __init__(self):
                 self.observation_space = Tuple(
                     [Discrete(5),
diff --git a/rllib/utils/actors.py b/rllib/utils/actors.py
index 70026d38f..92c7799ac 100644
--- a/rllib/utils/actors.py
+++ b/rllib/utils/actors.py
@@ -9,7 +9,7 @@ import ray
 logger = logging.getLogger(__name__)
 
 
-class TaskPool(object):
+class TaskPool:
     """Helper class for tracking the status of many in-flight actor tasks."""
 
     def __init__(self):
diff --git a/rllib/utils/debug.py b/rllib/utils/debug.py
index 2903d83c6..c38b466b4 100644
--- a/rllib/utils/debug.py
+++ b/rllib/utils/debug.py
@@ -106,7 +106,7 @@ def _summarize(obj):
         return obj
 
 
-class _StringValue(object):
+class _StringValue:
     def __init__(self, value):
         self.value = value
 
diff --git a/rllib/utils/filter.py b/rllib/utils/filter.py
index b0b27706a..f1b5ee081 100644
--- a/rllib/utils/filter.py
+++ b/rllib/utils/filter.py
@@ -9,7 +9,7 @@ import threading
 logger = logging.getLogger(__name__)
 
 
-class Filter(object):
+class Filter:
     """Processes input, possibly statefully."""
 
     def apply_changes(self, other, *args, **kwargs):
@@ -65,7 +65,7 @@ class NoFilter(Filter):
 
 
 # http://www.johndcook.com/blog/standard_deviation/
-class RunningStat(object):
+class RunningStat:
     def __init__(self, shape=None):
         self._n = 0
         self._M = np.zeros(shape)
diff --git a/rllib/utils/filter_manager.py b/rllib/utils/filter_manager.py
index d2362979a..1c7cc81c4 100644
--- a/rllib/utils/filter_manager.py
+++ b/rllib/utils/filter_manager.py
@@ -8,7 +8,7 @@ from ray.rllib.utils.memory import ray_get_and_free
 
 
 @DeveloperAPI
-class FilterManager(object):
+class FilterManager:
     """Manages filters and coordination across remote evaluators that expose
         `get_filters` and `sync_filters`.
     """
diff --git a/rllib/utils/policy_client.py b/rllib/utils/policy_client.py
index 3bfe73ff9..017f4fb39 100644
--- a/rllib/utils/policy_client.py
+++ b/rllib/utils/policy_client.py
@@ -19,7 +19,7 @@ except ImportError:
 
 
 @PublicAPI
-class PolicyClient(object):
+class PolicyClient:
     """REST client to interact with a RLlib policy server."""
 
     START_EPISODE = "START_EPISODE"
diff --git a/rllib/utils/schedules.py b/rllib/utils/schedules.py
index 12efda661..149402a2e 100644
--- a/rllib/utils/schedules.py
+++ b/rllib/utils/schedules.py
@@ -12,13 +12,13 @@ from __future__ import division
 from __future__ import print_function
 
 
-class Schedule(object):
+class Schedule:
     def value(self, t):
         """Value of the schedule at time t"""
         raise NotImplementedError()
 
 
-class ConstantSchedule(object):
+class ConstantSchedule:
     def __init__(self, value):
         """Value remains constant over time.
 
@@ -38,7 +38,7 @@ def linear_interpolation(l, r, alpha):
     return l + alpha * (r - l)
 
 
-class PiecewiseSchedule(object):
+class PiecewiseSchedule:
     def __init__(self,
                  endpoints,
                  interpolation=linear_interpolation,
@@ -82,7 +82,7 @@ class PiecewiseSchedule(object):
         return self._outside_value
 
 
-class LinearSchedule(object):
+class LinearSchedule:
     def __init__(self, schedule_timesteps, final_p, initial_p=1.0):
         """Linear interpolation between initial_p and final_p over
         schedule_timesteps. After this many timesteps pass final_p is
diff --git a/rllib/utils/tf_run_builder.py b/rllib/utils/tf_run_builder.py
index ed4525ddf..5521f364e 100644
--- a/rllib/utils/tf_run_builder.py
+++ b/rllib/utils/tf_run_builder.py
@@ -13,7 +13,7 @@ tf = try_import_tf()
 logger = logging.getLogger(__name__)
 
 
-class TFRunBuilder(object):
+class TFRunBuilder:
     """Used to incrementally build up a TensorFlow run.
 
     This is particularly useful for batching ops from multiple different
diff --git a/rllib/utils/timer.py b/rllib/utils/timer.py
index bfad20aab..e4be372b6 100644
--- a/rllib/utils/timer.py
+++ b/rllib/utils/timer.py
@@ -6,7 +6,7 @@ import numpy as np
 import time
 
 
-class TimerStat(object):
+class TimerStat:
     """A running stat for conveniently logging the duration of a code block.
 
     Example:
diff --git a/rllib/utils/window_stat.py b/rllib/utils/window_stat.py
index 21c93069a..a6277540b 100644
--- a/rllib/utils/window_stat.py
+++ b/rllib/utils/window_stat.py
@@ -5,7 +5,7 @@ from __future__ import print_function
 import numpy as np
 
 
-class WindowStat(object):
+class WindowStat:
     def __init__(self, name, n):
         self.name = name
         self.items = [None] * n
diff --git a/streaming/python/communication.py b/streaming/python/communication.py
index a5990165e..4aa2b32fe 100644
--- a/streaming/python/communication.py
+++ b/streaming/python/communication.py
@@ -31,7 +31,7 @@ def _hash(value):
         return int(hashlib.sha1(value).hexdigest(), 16)
 
 
-class DataChannel(object):
+class DataChannel:
     """A data channel for actor-to-actor communication.
 
     Attributes:
@@ -62,7 +62,7 @@ _CLOSE_FLAG = b" "
 
 
 # Pulls and merges data from multiple input channels
-class DataInput(object):
+class DataInput:
     """An input gate of an operator instance.
 
     The input gate pulls records from all input channels in a round-robin
@@ -124,7 +124,7 @@ class DataInput(object):
 
 
 # Selects output channel(s) and pushes data
-class DataOutput(object):
+class DataOutput:
     """An output gate of an operator instance.
 
     The output gate pushes records to output channels according to the
diff --git a/streaming/python/examples/key_selectors.py b/streaming/python/examples/key_selectors.py
index 4b1d2e7a3..6533d49b0 100644
--- a/streaming/python/examples/key_selectors.py
+++ b/streaming/python/examples/key_selectors.py
@@ -17,7 +17,7 @@ parser.add_argument("--input-file", required=True, help="the input text file")
 
 
 # A class used to check attribute-based key selection
-class Record(object):
+class Record:
     def __init__(self, record):
         k, _ = record
         self.word = k
diff --git a/streaming/python/examples/wordcount.py b/streaming/python/examples/wordcount.py
index 2062f2b7f..acf363f77 100644
--- a/streaming/python/examples/wordcount.py
+++ b/streaming/python/examples/wordcount.py
@@ -23,7 +23,7 @@ parser.add_argument(
 # A custom data source that reads articles from wikipedia
 # Custom data sources need to implement a get_next() method
 # that returns the next data element, in this case sentences
-class Wikipedia(object):
+class Wikipedia:
     def __init__(self, title_file):
         # Titles in this file will be as queries
         self.title_file = title_file
diff --git a/streaming/python/jobworker.py b/streaming/python/jobworker.py
index 03bfa1f97..b75e43420 100644
--- a/streaming/python/jobworker.py
+++ b/streaming/python/jobworker.py
@@ -16,7 +16,7 @@ logger = logging.getLogger(__name__)
 
 
 @ray.remote
-class JobWorker(object):
+class JobWorker:
     """A streaming job worker.
 
     Attributes:
diff --git a/streaming/python/operator.py b/streaming/python/operator.py
index cb698ff9a..ad2594f06 100644
--- a/streaming/python/operator.py
+++ b/streaming/python/operator.py
@@ -12,7 +12,7 @@ logger.setLevel("DEBUG")
 
 
 # Stream partitioning schemes
-class PScheme(object):
+class PScheme:
     def __init__(self, strategy, partition_fn=None):
         self.strategy = strategy
         self.partition_fn = partition_fn
@@ -51,7 +51,7 @@ class OpType(enum.Enum):
 
 
 # A logical dataflow operator
-class Operator(object):
+class Operator:
     def __init__(self,
                  id,
                  op_type,
diff --git a/streaming/python/streaming.py b/streaming/python/streaming.py
index 6f93322bf..32d2d1a02 100644
--- a/streaming/python/streaming.py
+++ b/streaming/python/streaming.py
@@ -34,7 +34,7 @@ all_to_all_strategies = [
 
 
 # Environment configuration
-class Conf(object):
+class Conf:
     """Environment configuration.
 
     This class includes all information about the configuration of the
@@ -253,7 +253,7 @@ class ExecutionGraph:
 
 
 # The execution environment for a streaming job
-class Environment(object):
+class Environment:
     """A streaming environment.
 
     This class is responsible for constructing the logical and the
@@ -394,7 +394,7 @@ class Environment(object):
 
 
 # A DataStream corresponds to an edge in the logical dataflow
-class DataStream(object):
+class DataStream:
     """A data stream.
 
     This class contains all information about a logical stream, i.e. an edge