diff --git a/stable_baselines3/cmaes/policies.py b/stable_baselines3/cmaes/policies.py
index 09ca548..98bcd10 100644
--- a/stable_baselines3/cmaes/policies.py
+++ b/stable_baselines3/cmaes/policies.py
@@ -1,4 +1,4 @@
-from typing import Any, Dict, List, Optional, Type, Union
+from typing import Any, Dict, List, Optional, Type
 
 import gym
 import torch as th
@@ -23,7 +23,6 @@ class ActorCMA(BasePolicy):
     :param activation_fn: (Type[nn.Module]) Activation function
     :param normalize_images: (bool) Whether to normalize images or not,
          dividing by 255.0 (True by default)
-    :param device: (Union[th.device, str]) Device on which the code should run.
     """
 
     def __init__(
@@ -35,14 +34,12 @@ class ActorCMA(BasePolicy):
         features_dim: int,
         activation_fn: Type[nn.Module] = nn.ReLU,
         normalize_images: bool = True,
-        device: Union[th.device, str] = "auto",
     ):
         super(ActorCMA, self).__init__(
             observation_space,
             action_space,
             features_extractor=features_extractor,
             normalize_images=normalize_images,
-            device=device,
             squash_output=True,
         )
 
@@ -85,7 +82,6 @@ class CMAESPolicy(BasePolicy):
         observation_space: gym.spaces.Space,
         action_space: gym.spaces.Space,
         net_arch: Optional[List[int]] = None,
-        device: Union[th.device, str] = "auto",
         activation_fn: Type[nn.Module] = nn.ReLU,
         features_extractor_class: Type[BaseFeaturesExtractor] = FlattenExtractor,
         features_extractor_kwargs: Optional[Dict[str, Any]] = None,
@@ -96,7 +92,6 @@ class CMAESPolicy(BasePolicy):
         super(CMAESPolicy, self).__init__(
             observation_space,
             action_space,
-            device,
             features_extractor_class,
             features_extractor_kwargs,
             optimizer_class=optimizer_class,
@@ -124,7 +119,6 @@ class CMAESPolicy(BasePolicy):
             "net_arch": self.net_arch,
             "activation_fn": self.activation_fn,
             "normalize_images": normalize_images,
-            "device": device,
             "use_sde": True,
         }
         self.actor = Actor(**self.net_args).to(self.device)
@@ -161,7 +155,6 @@ class CnnPolicy(CMAESPolicy):
         observation_space: gym.spaces.Space,
         action_space: gym.spaces.Space,
         net_arch: Optional[List[int]] = None,
-        device: Union[th.device, str] = "auto",
         activation_fn: Type[nn.Module] = nn.ReLU,
         features_extractor_class: Type[BaseFeaturesExtractor] = NatureCNN,
         features_extractor_kwargs: Optional[Dict[str, Any]] = None,
@@ -173,7 +166,6 @@ class CnnPolicy(CMAESPolicy):
             observation_space,
             action_space,
             net_arch,
-            device,
             activation_fn,
             features_extractor_class,
             features_extractor_kwargs,
diff --git a/stable_baselines3/tqc/policies.py b/stable_baselines3/tqc/policies.py
index 61a30b4..52e75f8 100644
--- a/stable_baselines3/tqc/policies.py
+++ b/stable_baselines3/tqc/policies.py
@@ -1,4 +1,4 @@
-from typing import Any, Callable, Dict, List, Optional, Tuple, Type, Union
+from typing import Any, Callable, Dict, List, Optional, Tuple, Type
 
 import gym
 import torch as th
@@ -38,7 +38,6 @@ class Actor(BasePolicy):
     :param clip_mean: (float) Clip the mean output when using gSDE to avoid numerical instability.
     :param normalize_images: (bool) Whether to normalize images or not,
          dividing by 255.0 (True by default)
-    :param device: (Union[th.device, str]) Device on which the code should run.
     """
 
     def __init__(
@@ -56,14 +55,12 @@ class Actor(BasePolicy):
         use_expln: bool = False,
         clip_mean: float = 2.0,
         normalize_images: bool = True,
-        device: Union[th.device, str] = "auto",
     ):
         super(Actor, self).__init__(
             observation_space,
             action_space,
             features_extractor=features_extractor,
             normalize_images=normalize_images,
-            device=device,
             squash_output=True,
         )
 
@@ -213,7 +210,6 @@ class Critic(BaseModel):
     :param activation_fn: (Type[nn.Module]) Activation function
     :param normalize_images: (bool) Whether to normalize images or not,
          dividing by 255.0 (True by default)
-    :param device: (Union[th.device, str]) Device on which the code should run.
     """
 
     def __init__(
@@ -225,16 +221,14 @@ class Critic(BaseModel):
         features_dim: int,
         activation_fn: Type[nn.Module] = nn.ReLU,
         normalize_images: bool = True,
-        device: Union[th.device, str] = "auto",
         n_quantiles: int = 25,
         n_critics: int = 2,
     ):
-        super(Critic, self).__init__(
+        super().__init__(
             observation_space,
             action_space,
             features_extractor=features_extractor,
             normalize_images=normalize_images,
-            device=device,
         )
 
         action_dim = get_action_dim(self.action_space)
@@ -268,7 +262,6 @@ class TQCPolicy(BasePolicy):
     :param action_space: (gym.spaces.Space) Action space
     :param lr_schedule: (callable) Learning rate schedule (could be constant)
     :param net_arch: (Optional[List[int]]) The specification of the policy and value networks.
-    :param device: (str or th.device) Device on which the code should run.
     :param activation_fn: (Type[nn.Module]) Activation function
     :param use_sde: (bool) Whether to use State Dependent Exploration or not
     :param log_std_init: (float) Initial value for the log standard deviation
@@ -296,7 +289,6 @@ class TQCPolicy(BasePolicy):
         action_space: gym.spaces.Space,
         lr_schedule: Callable,
         net_arch: Optional[List[int]] = None,
-        device: Union[th.device, str] = "auto",
         activation_fn: Type[nn.Module] = nn.ReLU,
         use_sde: bool = False,
         log_std_init: float = -3,
@@ -314,7 +306,6 @@ class TQCPolicy(BasePolicy):
         super(TQCPolicy, self).__init__(
             observation_space,
             action_space,
-            device,
             features_extractor_class,
             features_extractor_kwargs,
             optimizer_class=optimizer_class,
@@ -342,7 +333,6 @@ class TQCPolicy(BasePolicy):
             "net_arch": self.net_arch,
             "activation_fn": self.activation_fn,
             "normalize_images": normalize_images,
-            "device": device,
         }
         self.actor_kwargs = self.net_args.copy()
         self.critic_kwargs = self.net_args.copy()
@@ -427,7 +417,6 @@ class CnnPolicy(TQCPolicy):
     :param action_space: (gym.spaces.Space) Action space
     :param lr_schedule: (callable) Learning rate schedule (could be constant)
     :param net_arch: (Optional[List[int]]) The specification of the policy and value networks.
-    :param device: (str or th.device) Device on which the code should run.
     :param activation_fn: (Type[nn.Module]) Activation function
     :param use_sde: (bool) Whether to use State Dependent Exploration or not
     :param log_std_init: (float) Initial value for the log standard deviation
@@ -453,7 +442,6 @@ class CnnPolicy(TQCPolicy):
         action_space: gym.spaces.Space,
         lr_schedule: Callable,
         net_arch: Optional[List[int]] = None,
-        device: Union[th.device, str] = "auto",
         activation_fn: Type[nn.Module] = nn.ReLU,
         use_sde: bool = False,
         log_std_init: float = -3,
@@ -473,7 +461,6 @@ class CnnPolicy(TQCPolicy):
             action_space,
             lr_schedule,
             net_arch,
-            device,
             activation_fn,
             use_sde,
             log_std_init,