Fix graph overlaps and some typos. (#16)

* Minor typo fixes for 2_gym_wrappers_saving_loading.ipynb

* Adjust plot spacing and fix some typos.

* The first plot cell defines a nice spacing to use which ensures
  that the y axis of the second plot doesn't overlap the first plot.
  This just copies that block of code to the other plotting cells.

* Extract plotting into a helper function.

* Clean up plot util function.

* Undo unrelated colab changes.

* Fix my own typo.

Author: dan-pandori

2_gym_wrappers_saving_loading.ipynb

@@ -378,7 +378,7 @@
       "source": [
         "## Second example: normalize actions\n",
         "\n",
-        "It is usually a good idea to normalize observations and actions before giving it to the agent, this prevent [hard to debug issue](https://github.com/hill-a/stable-baselines/issues/473).\n",
+        "It is usually a good idea to normalize observations and actions before giving it to the agent, this prevents this [hard to debug issue](https://github.com/hill-a/stable-baselines/issues/473).\n",
         "\n",
         "In this example, we are going to normalize the action space of *Pendulum-v1* so it lies in [-1, 1] instead of [-2, 2].\n",
         "\n",
@@ -425,7 +425,6 @@
         "    \"\"\"\n",
         "    Reset the environment \n",
         "    \"\"\"\n",
-        "    # Reset the counter\n",
         "    return self.env.reset()\n",
         "\n",
         "  def step(self, action):\n",
@@ -505,7 +504,7 @@
       "source": [
         "#### Test with a RL algorithm\n",
         "\n",
-        "We are going to use the Monitor wrapper of stable baselines, wich allow to monitor training stats (mean episode reward, mean episode length)"
+        "We are going to use the Monitor wrapper of stable baselines, which allow to monitor training stats (mean episode reward, mean episode length)"
       ]
     },
     {
@@ -610,7 +609,7 @@
       "source": [
         "## Additional wrappers: VecEnvWrappers\n",
         "\n",
-        "In the same vein as gym wrappers, stable baselines provide wrappers for `VecEnv`. Among the different that exist (and you can create your own), you should know: \n",
+        "In the same vein as gym wrappers, stable baselines provide wrappers for `VecEnv`. Among the different wrappers that exist (and you can create your own), you should know: \n",
         "\n",
         "- VecNormalize: it computes a running mean and standard deviation to normalize observation and returns\n",
         "- VecFrameStack: it stacks several consecutive observations (useful to integrate time in the observation, e.g. sucessive frame of an atari game)\n",
@@ -760,7 +759,7 @@
         "\n",
         "# Reset the environment\n",
         "\n",
-        "# Take random actions in the enviromnent and check\n",
+        "# Take random actions in the environment and check\n",
         "# that it returns the correct values after the end of each episode\n",
         "\n",
         "# ====================== #"
@@ -851,7 +850,7 @@
         "        time_feature = 1 - (self._current_step / self._max_steps)\n",
         "        if self._test_mode:\n",
         "            time_feature = 1.0\n",
-        "        # Optionnaly: concatenate [time_feature, time_feature ** 2]\n",
+        "        # Optionally: concatenate [time_feature, time_feature ** 2]\n",
         "        return np.concatenate((obs, [time_feature]))"
       ],
       "execution_count": 0,