diff --git a/kernel_tuner/runners/parallel.py b/kernel_tuner/runners/parallel.py
index e8134116..871b9322 100644
--- a/kernel_tuner/runners/parallel.py
+++ b/kernel_tuner/runners/parallel.py
@@ -144,6 +144,9 @@ def _setup_tuning_options(self, tuning_options, evaluations_per_strategy):
         return new_tuning_options
     
     def _process_results_ensemble(self, all_results):
+        """
+        Process the results from the ensemble execution.
+        """
         results = []
         tuning_options_list = []
 
@@ -155,6 +158,9 @@ def _process_results_ensemble(self, all_results):
 
 
     def parallel_function_evaluation(self, tuning_options, parameter_space):
+        """
+        Perform parallel function evaluation.
+        """
         # Create a pool of RemoteActor actors
         self.actor_pool = ActorPool(self.actors)
         # Distribute execution of the `execute` method across the actor pool with varying parameters and tuning options, collecting the results asynchronously.
@@ -164,6 +170,9 @@ def parallel_function_evaluation(self, tuning_options, parameter_space):
         return results, tuning_options_list
     
     def _process_results(self, all_results, searchspace):
+        """
+        Process the results and remove duplicates based on the searchspace.
+        """
         unique_configs = set()
         final_results = []
 
@@ -176,12 +185,18 @@ def _process_results(self, all_results, searchspace):
         return final_results
     
     def _calculate_simulated_time(self, tuning_options_list):
+        """
+        Calculate the maximum simulated time from the list of tuning options.
+        """
         simulated_times = []
         for tuning_options in tuning_options_list:
             simulated_times.append(tuning_options.simulated_time)
         return max(simulated_times)
 
     def _check_gpus_equals(self):
+        """
+        Check if all GPUs are of the same type.
+        """
         gpu_types = []
         env_refs = [actor.get_environment.remote() for actor in self.actors]
         environments = ray.get(env_refs)
@@ -194,6 +209,9 @@ def _check_gpus_equals(self):
             return False
 
     def clean_up_ray(self):
+        """
+        Clean up Ray actors and cache manager.
+        """
         if self.actors is not None:
             for actor in self.actors:
                 ray.kill(actor)
diff --git a/kernel_tuner/strategies/ensemble.py b/kernel_tuner/strategies/ensemble.py
index 9cdc0b90..7e66f036 100644
--- a/kernel_tuner/strategies/ensemble.py
+++ b/kernel_tuner/strategies/ensemble.py
@@ -50,19 +50,23 @@ def tune(searchspace: Searchspace, runner, tuning_options, cache_manager=None, a
     ensemble = options.get('ensemble', ["diff_evo", "diff_evo"])
     ensemble_size = len(ensemble)
 
+    # setup strategy options
     if 'bayes_opt' in ensemble: # All strategies start from a random sample except for BO
         tuning_options.strategy_options["samplingmethod"] = 'random'
     tuning_options.strategy_options["max_fevals"] = options.get("max_fevals", 100 * ensemble_size)
     tuning_options.strategy_options['check_and_retrieve'] = True
 
+    # define number of ray actors needed
     if num_devices < ensemble_size:
         warnings.warn("Number of devices is less than the number of strategies in the ensemble. Some strategies will wait until devices are available.", UserWarning)
     num_actors = num_devices if ensemble_size > num_devices else ensemble_size
 
     ensemble = [strategy_map[strategy] for strategy in ensemble]
+
     parallel_runner = ParallelRunner(runner.kernel_source, runner.kernel_options, runner.device_options, 
                                     runner.iterations, runner.observers, num_gpus=num_actors, cache_manager=cache_manager,
                                     simulation_mode=simulation_mode, actors=actors)
+    
     final_results = parallel_runner.run(tuning_options=tuning_options, ensemble=ensemble, searchspace=searchspace)
 
     if clean_up: