diff --git a/pl_bolts/models/rl/common/gym_wrappers.py b/pl_bolts/models/rl/common/gym_wrappers.py
index 37f8ee5b50..cbf95bcbdc 100644
--- a/pl_bolts/models/rl/common/gym_wrappers.py
+++ b/pl_bolts/models/rl/common/gym_wrappers.py
@@ -4,21 +4,28 @@
 """
 import collections
 
-import gym
-import gym.spaces
 import numpy as np
 import torch
 
-from pl_bolts.utils import _OPENCV_AVAILABLE
+from pl_bolts.utils import _GYM_AVAILABLE, _OPENCV_AVAILABLE
 from pl_bolts.utils.warnings import warn_missing_pkg
 
+if _GYM_AVAILABLE:
+    import gym.spaces
+    from gym import ObservationWrapper, Wrapper
+    from gym import make as gym_make
+else:  # pragma: no-cover
+    warn_missing_pkg('gym')
+    Wrapper = object
+    ObservationWrapper = object
+
 if _OPENCV_AVAILABLE:
     import cv2
 else:
     warn_missing_pkg('cv2', pypi_name='opencv-python')  # pragma: no-cover
 
 
-class ToTensor(gym.Wrapper):
+class ToTensor(Wrapper):
     """For environments where the user need to press FIRE for the game to start."""
 
     def __init__(self, env=None):
@@ -34,7 +41,7 @@ def reset(self):
         return torch.tensor(self.env.reset())
 
 
-class FireResetEnv(gym.Wrapper):
+class FireResetEnv(Wrapper):
     """For environments where the user need to press FIRE for the game to start."""
 
     def __init__(self, env=None):
@@ -58,7 +65,7 @@ def reset(self):
         return obs
 
 
-class MaxAndSkipEnv(gym.Wrapper):
+class MaxAndSkipEnv(Wrapper):
     """Return only every `skip`-th frame"""
 
     def __init__(self, env=None, skip=4):
@@ -88,7 +95,7 @@ def reset(self):
         return obs
 
 
-class ProcessFrame84(gym.ObservationWrapper):
+class ProcessFrame84(ObservationWrapper):
     """preprocessing images from env"""
 
     def __init__(self, env=None):
@@ -121,7 +128,7 @@ def process(frame):
         return x_t.astype(np.uint8)
 
 
-class ImageToPyTorch(gym.ObservationWrapper):
+class ImageToPyTorch(ObservationWrapper):
     """converts image to pytorch format"""
 
     def __init__(self, env):
@@ -142,7 +149,7 @@ def observation(observation):
         return np.moveaxis(observation, 2, 0)
 
 
-class ScaledFloatFrame(gym.ObservationWrapper):
+class ScaledFloatFrame(ObservationWrapper):
     """scales the pixels"""
 
     @staticmethod
@@ -150,7 +157,7 @@ def observation(obs):
         return np.array(obs).astype(np.float32) / 255.0
 
 
-class BufferWrapper(gym.ObservationWrapper):
+class BufferWrapper(ObservationWrapper):
     """"Wrapper for image stacking"""
 
     def __init__(self, env, n_steps, dtype=np.float32):
@@ -176,7 +183,7 @@ def observation(self, observation):
         return self.buffer
 
 
-class DataAugmentation(gym.ObservationWrapper):
+class DataAugmentation(ObservationWrapper):
     """
     Carries out basic data augmentation on the env observations
     - ToTensor
@@ -197,7 +204,7 @@ def observation(self, obs):
 
 def make_environment(env_name):
     """Convert environment with wrappers"""
-    env = gym.make(env_name)
+    env = gym_make(env_name)
     env = MaxAndSkipEnv(env)
     env = FireResetEnv(env)
     env = ProcessFrame84(env)
diff --git a/pl_bolts/models/rl/dqn_model.py b/pl_bolts/models/rl/dqn_model.py
index 0635ead24c..6f5e958a36 100644
--- a/pl_bolts/models/rl/dqn_model.py
+++ b/pl_bolts/models/rl/dqn_model.py
@@ -18,15 +18,17 @@
 from pl_bolts.datamodules.experience_source import Experience, ExperienceSourceDataset
 from pl_bolts.losses.rl import dqn_loss
 from pl_bolts.models.rl.common.agents import ValueAgent
+from pl_bolts.models.rl.common.gym_wrappers import make_environment
 from pl_bolts.models.rl.common.memory import MultiStepBuffer
 from pl_bolts.models.rl.common.networks import CNN
 from pl_bolts.utils import _GYM_AVAILABLE
 from pl_bolts.utils.warnings import warn_missing_pkg
 
 if _GYM_AVAILABLE:
-    from pl_bolts.models.rl.common.gym_wrappers import gym, make_environment
+    from gym import Env
 else:
     warn_missing_pkg('gym')  # pragma: no-cover
+    Env = object
 
 
 class DQN(pl.LightningModule):
@@ -336,7 +338,7 @@ def test_dataloader(self) -> DataLoader:
         return self._dataloader()
 
     @staticmethod
-    def make_environment(env_name: str, seed: Optional[int] = None) -> gym.Env:
+    def make_environment(env_name: str, seed: Optional[int] = None) -> Env:
         """
         Initialise gym  environment