diff --git a/tests/tpot_tests.py b/tests/tpot_tests.py
index 0088833c..d8f878e8 100644
--- a/tests/tpot_tests.py
+++ b/tests/tpot_tests.py
@@ -964,12 +964,39 @@ def test_fit_4():
     tpot_obj.generations == 20
 
     tpot_obj.fit(training_features, training_target)
-
+    assert tpot_obj._pop == []
     assert isinstance(tpot_obj._optimized_pipeline, creator.Individual)
     assert not (tpot_obj._start_datetime is None)
 
 
 def test_fit_5():
+    """Assert that the TPOT fit function provides an optimized pipeline with max_time_mins of 2 second with warm_start=True."""
+    tpot_obj = TPOTClassifier(
+        random_state=42,
+        population_size=2,
+        generations=None,
+        verbosity=0,
+        max_time_mins=3/60.,
+        config_dict='TPOT light',
+        warm_start=True
+    )
+    tpot_obj._fit_init()
+    assert tpot_obj.generations == 1000000
+
+    # reset generations to 20 just in case that the failed test may take too much time
+    tpot_obj.generations == 20
+
+    tpot_obj.fit(training_features, training_target)
+    assert tpot_obj._pop != []
+    assert isinstance(tpot_obj._optimized_pipeline, creator.Individual)
+    assert not (tpot_obj._start_datetime is None)
+    # rerun it
+    tpot_obj.fit(training_features, training_target)
+    assert tpot_obj._pop != []
+
+
+
+def test_fit_6():
     """Assert that the TPOT fit function provides an optimized pipeline with pandas DataFrame"""
     tpot_obj = TPOTClassifier(
         random_state=42,
diff --git a/tpot/base.py b/tpot/base.py
index 7bfc56e5..9b003ce7 100644
--- a/tpot/base.py
+++ b/tpot/base.py
@@ -752,10 +752,6 @@ def pareto_eq(ind1, ind2):
                     per_generation_function=self._check_periodic_pipeline
                 )
 
-            # store population for the next call
-            if self.warm_start:
-                self._pop = pop
-
         # Allow for certain exceptions to signal a premature fit() cancellation
         except (KeyboardInterrupt, SystemExit, StopIteration) as e:
             if self.verbosity > 0:
@@ -763,6 +759,9 @@ def pareto_eq(ind1, ind2):
                 self._pbar.write('{}\nTPOT closed prematurely. Will use the current best pipeline.'.format(e),
                                  file=self._file)
         finally:
+            # clean population for the next call if warm_start=False
+            if not self.warm_start:
+                self._pop = []
             # keep trying 10 times in case weird things happened like multiple CTRL+C or exceptions
             attempts = 10
             for attempt in range(attempts):
@@ -1383,6 +1382,8 @@ def _evaluate_individuals(self, population, features, target, sample_weight=None
                 ind.fitness.values = (5000.,-float('inf'))
 
             self._pareto_front.update(population)
+            
+            self._pop = population
             raise KeyboardInterrupt
 
         self._update_evaluated_individuals_(result_score_list, eval_individuals_str, operator_counts, stats_dicts)