diff --git a/docs/misc/changelog.rst b/docs/misc/changelog.rst
index 3f6e087..e86c8e2 100644
--- a/docs/misc/changelog.rst
+++ b/docs/misc/changelog.rst
@@ -20,6 +20,8 @@ Bug Fixes:
 - Fix ignoring the exclude parameter when recording logs using json, csv or log as logging format (@SwamyDev)
 - Make ``make_vec_env`` support the ``env_kwargs`` argument when using an env ID str (@ManifoldFR)
 - Fix model creation initializing CUDA even when `device="cpu"` is provided
+- Fix ``check_env`` not checking if the env has a Dict actionspace before calling ``_check_nan`` (@wmmc88)
+- Update the check for spaces unsupported by Stable Baselines 3 to include checks on the action space (@wmmc88)
 
 Deprecations:
 ^^^^^^^^^^^^^
@@ -457,4 +459,4 @@ And all the contributors:
 @MarvineGothic @jdossgollin @SyllogismRXS @rusu24edward @jbulow @Antymon @seheevic @justinkterry @edbeeching
 @flodorner @KuKuXia @NeoExtended @PartiallyTyped @mmcenta @richardwu @kinalmehta @rolandgvc @tkelestemur @mloo3
 @tirafesi @blurLake @koulakis @joeljosephjin @shwang @rk37 @andyshih12 @RaphaelWag @xicocaio
-@diditforlulz273 @liorcohen5 @ManifoldFR @mloo3 @SwamyDev
+@diditforlulz273 @liorcohen5 @ManifoldFR @mloo3 @SwamyDev @wmmc88
diff --git a/stable_baselines3/common/env_checker.py b/stable_baselines3/common/env_checker.py
index 4a4d075..262722f 100644
--- a/stable_baselines3/common/env_checker.py
+++ b/stable_baselines3/common/env_checker.py
@@ -8,12 +8,12 @@ from gym import spaces
 from stable_baselines3.common.vec_env import DummyVecEnv, VecCheckNan
 
 
-def _enforce_array_obs(observation_space: spaces.Space) -> bool:
+def _is_numpy_array_space(space: spaces.Space) -> bool:
     """
-    Whether to check that the returned observation is a numpy array
-    it is not mandatory for `Dict` and `Tuple` spaces.
+    Returns False if provided space is not representable as a single numpy array
+    (e.g. Dict and Tuple spaces return False)
     """
-    return not isinstance(observation_space, (spaces.Dict, spaces.Tuple))
+    return not isinstance(space, (spaces.Dict, spaces.Tuple))
 
 
 def _check_image_input(observation_space: spaces.Box) -> None:
@@ -45,8 +45,8 @@ def _check_image_input(observation_space: spaces.Box) -> None:
         )
 
 
-def _check_unsupported_obs_spaces(env: gym.Env, observation_space: spaces.Space) -> None:
-    """Emit warnings when the observation space used is not supported by Stable-Baselines."""
+def _check_unsupported_spaces(env: gym.Env, observation_space: spaces.Space, action_space: spaces.Space) -> None:
+    """Emit warnings when the observation space or action space used is not supported by Stable-Baselines."""
 
     if isinstance(observation_space, spaces.Dict) and not isinstance(env, gym.GoalEnv):
         warnings.warn(
@@ -65,6 +65,13 @@ def _check_unsupported_obs_spaces(env: gym.Env, observation_space: spaces.Space)
             "you will need to flatten the observation and maybe use a custom policy. "
         )
 
+    if not _is_numpy_array_space(action_space):
+        warnings.warn(
+            "The action space is not based off a numpy array. Typically this means it's either a Dict or Tuple space. "
+            "This type of action space is currently not supported by Stable Baselines 3. You should try to flatten the "
+            "action using a wrapper."
+        )
+
 
 def _check_nan(env: gym.Env) -> None:
     """Check for Inf and NaN using the VecWrapper."""
@@ -87,7 +94,7 @@ def _check_obs(obs: Union[tuple, dict, np.ndarray, int], observation_space: spac
     # The check for a GoalEnv is done by the base class
     if isinstance(observation_space, spaces.Discrete):
         assert isinstance(obs, int), "The observation returned by `{}()` method must be an int".format(method_name)
-    elif _enforce_array_obs(observation_space):
+    elif _is_numpy_array_space(observation_space):
         assert isinstance(obs, np.ndarray), "The observation returned by `{}()` method must be a numpy array".format(
             method_name
         )
@@ -201,7 +208,7 @@ def check_env(env: gym.Env, warn: bool = True, skip_render_check: bool = True) -
     # Warn the user if needed.
     # A warning means that the environment may run but not work properly with Stable Baselines algorithms
     if warn:
-        _check_unsupported_obs_spaces(env, observation_space)
+        _check_unsupported_spaces(env, observation_space, action_space)
 
         # If image, check the low and high values, the type and the number of channels
         # and the shape (minimal value)
@@ -234,5 +241,5 @@ def check_env(env: gym.Env, warn: bool = True, skip_render_check: bool = True) -
         _check_render(env, warn=warn)
 
     # The check only works with numpy arrays
-    if _enforce_array_obs(observation_space):
+    if _is_numpy_array_space(observation_space) and _is_numpy_array_space(action_space):
         _check_nan(env)
diff --git a/tests/test_env_checker.py b/tests/test_env_checker.py
new file mode 100644
index 0000000..6364bd4
--- /dev/null
+++ b/tests/test_env_checker.py
@@ -0,0 +1,31 @@
+import gym
+import numpy as np
+import pytest
+from gym.spaces import Box, Dict, Discrete
+
+from stable_baselines3.common.env_checker import check_env
+
+
+class ActionDictTestEnv(gym.Env):
+    action_space = Dict({"position": Discrete(1), "velocity": Discrete(1)})
+    observation_space = Box(low=-1.0, high=2.0, shape=(3,), dtype=np.float32)
+
+    def step(self, action):
+        observation = np.array([1.0, 1.5, 0.5])
+        reward = 1
+        done = True
+        info = {}
+        return observation, reward, done, info
+
+    def reset(self):
+        return np.array([1.0, 1.5, 0.5])
+
+    def render(self, mode="human"):
+        pass
+
+
+def test_check_env_dict_action():
+    test_env = ActionDictTestEnv()
+
+    with pytest.warns(Warning):
+        check_env(env=test_env, warn=True)