Merge pull request #338 from KernelTuner/bo-remove-searchspace-fully-observed-check

Remove `error_message_searchspace_fully_observed` message in BO

Author: stijnh

kernel_tuner/strategies/bayes_opt.py

@@ -268,7 +268,6 @@ def get_hyperparam(name: str, default, supported_values=list()):
         self.__valid_observations = list()
         self.unvisited_cache = self.unvisited()
         time_setup = time.perf_counter_ns()
-        self.error_message_searchspace_fully_observed = "The search space has been fully observed"
 
         # take initial sample
         if self.num_initial_samples > 0:
@@ -572,7 +571,7 @@ def __optimize(self, max_fevals):
         """Find the next best candidate configuration(s), evaluate those and update the model accordingly."""
         while self.fevals < max_fevals:
             if self.__visited_num >= self.searchspace_size:
-                raise ValueError(self.error_message_searchspace_fully_observed)
+                break
             predictions, _, std = self.predict_list(self.unvisited_cache)
             hyperparam = self.contextual_variance(std)
             list_of_acquisition_values = self.__af(predictions, hyperparam)
@@ -611,7 +610,7 @@ def __optimize_multi(self, max_fevals):
             predictions, _, std = self.predict_list(self.unvisited_cache)
             hyperparam = self.contextual_variance(std)
             if self.__visited_num >= self.searchspace_size:
-                raise ValueError(self.error_message_searchspace_fully_observed)
+                break
             time_predictions = time.perf_counter_ns()
             actual_candidate_params = list()
             actual_candidate_indices = list()
@@ -727,7 +726,7 @@ def __optimize_multi_advanced(self, max_fevals, increase_precision=False):
             if single_af:
                 return self.__optimize(max_fevals)
             if self.__visited_num >= self.searchspace_size:
-                raise ValueError(self.error_message_searchspace_fully_observed)
+                break
             observations_median = np.median(self.__valid_observations)
             if increase_precision is False:
                 predictions, _, std = self.predict_list(self.unvisited_cache)
@@ -835,7 +834,7 @@ def __optimize_multi_fast(self, max_fevals):
             predictions, _, std = self.predict_list(self.unvisited_cache)
             hyperparam = self.contextual_variance(std)
             if self.__visited_num >= self.searchspace_size:
-                raise ValueError(self.error_message_searchspace_fully_observed)
+                break
             for af in aqfs:
                 if self.__visited_num >= self.searchspace_size or self.fevals >= max_fevals:
                     break
@@ -876,7 +875,7 @@ def __optimize_multi_ultrafast(self, max_fevals, predict_eval_ratio=5):
             eval_start = time.perf_counter()
             hyperparam = self.contextual_variance(std)
             if self.__visited_num >= self.searchspace_size:
-                raise ValueError(self.error_message_searchspace_fully_observed)
+                break
             for af in aqfs:
                 if self.__visited_num >= self.searchspace_size or self.fevals >= max_fevals:
                     break