Support constant_liar in multi-objective TPESampler

optuna/samplers/_tpe/sampler.py

@@ -205,10 +205,6 @@ def objective(trial):
                 and the durations of the running states are significant, and/or the number of
                 workers is high.
 
-            .. note::
-                This feature can be used for only single-objective optimization; this argument is
-                ignored for multi-objective optimization.
-
             .. note::
                 Added in v2.8.0 as an experimental feature. The interface may change in newer
                 versions without prior notice. See
@@ -437,7 +433,7 @@ def _get_internal_repr(
     def _sample(
         self, study: Study, trial: FrozenTrial, search_space: Dict[str, BaseDistribution]
     ) -> Dict[str, Any]:
-        if self._constant_liar and not study._is_multi_objective():
+        if self._constant_liar:
             states = [TrialState.COMPLETE, TrialState.PRUNED, TrialState.RUNNING]
         else:
             states = [TrialState.COMPLETE, TrialState.PRUNED]

tests/samplers_tests/tpe_tests/test_sampler.py

@@ -1,3 +1,5 @@
+from __future__ import annotations
+
 import random
 from typing import Callable
 from typing import Dict
@@ -817,6 +819,62 @@ def test_split_trials(direction: str, constant_liar: bool, constraints: bool) ->
         assert above_trial_numbers == list(range(n_below, len(trials)))
 
 
+@pytest.mark.parametrize(
+    "directions", [["minimize", "minimize"], ["maximize", "maximize"], ["minimize", "maximize"]]
+)
+def test_split_trials_for_multiobjective_constant_liar(directions: list[str]) -> None:
+    study = optuna.create_study(directions=directions)
+    for obj1 in [-float("inf"), 0, 1, float("inf")]:
+        val1 = obj1 if directions[0] == "minimize" else -obj1
+        for obj2 in [-float("inf"), 0, 1, float("inf")]:
+            val2 = obj2 if directions[1] == "minimize" else -obj2
+            study.add_trial(
+                optuna.create_trial(
+                    state=optuna.trial.TrialState.COMPLETE,
+                    values=[val1, val2],
+                    params={"x": 0},
+                    distributions={"x": optuna.distributions.FloatDistribution(-1.0, 1.0)},
+                )
+            )
+
+    for _ in range(5):
+        study.add_trial(
+            optuna.create_trial(
+                state=optuna.trial.TrialState.RUNNING,
+                params={"x": 0},
+                distributions={"x": optuna.distributions.FloatDistribution(-1.0, 1.0)},
+            )
+        )
+
+    study.add_trial(
+        optuna.create_trial(
+            state=optuna.trial.TrialState.FAIL,
+        )
+    )
+
+    study.add_trial(
+        optuna.create_trial(
+            state=optuna.trial.TrialState.WAITING,
+        )
+    )
+
+    states = [optuna.trial.TrialState.COMPLETE, optuna.trial.TrialState.RUNNING]
+    trials = study.get_trials(states=states)
+    finished_trials = study.get_trials(states=(optuna.trial.TrialState.COMPLETE,))
+    ground_truth = [0, 1, 4, 2, 8, 5, 3, 6, 9, 12, 7, 10, 13, 11, 14, 15, 16, 17, 18, 19, 20]
+    for n_below in range(1, len(finished_trials) + 1):
+        below_trials, above_trials = _tpe.sampler._split_trials(
+            study,
+            trials,
+            n_below,
+            constraints_enabled=False,
+        )
+        below_trial_numbers = [trial.number for trial in below_trials]
+        assert below_trial_numbers == np.sort(ground_truth[:n_below]).tolist()
+        above_trial_numbers = [trial.number for trial in above_trials]
+        assert above_trial_numbers == np.sort(ground_truth[n_below:]).tolist()
+
+
 @pytest.mark.parametrize("direction", ["minimize", "maximize"])
 def test_split_complete_trials_single_objective(direction: str) -> None:
     study = optuna.create_study(direction=direction)
@@ -1052,7 +1110,8 @@ def test_constant_liar_experimental_warning() -> None:
 
 
 @pytest.mark.parametrize("multivariate", [True, False])
-def test_constant_liar_with_running_trial(multivariate: bool) -> None:
+@pytest.mark.parametrize("multiobjective", [True, False])
+def test_constant_liar_with_running_trial(multivariate: bool, multiobjective: bool) -> None:
     with warnings.catch_warnings():
         warnings.simplefilter("ignore", optuna.exceptions.ExperimentalWarning)
         sampler = TPESampler(multivariate=multivariate, constant_liar=True, n_startup_trials=0)
@@ -1064,7 +1123,7 @@ def test_constant_liar_with_running_trial(multivariate: bool) -> None:
     trial0.suggest_int("x", 0, 10)
     trial0.suggest_float("y", 0, 10)
     trial0.suggest_categorical("z", [0, 1, 2])
-    study.tell(trial0, 0)
+    study.tell(trial0, [0, 0] if multiobjective else 0)
 
     # Add running trials.
     trial1 = study.ask()
@@ -1079,7 +1138,7 @@ def test_constant_liar_with_running_trial(multivariate: bool) -> None:
     trial.suggest_int("x", 0, 10)
     trial.suggest_float("y", 0, 10)
     trial.suggest_categorical("z", [0, 1, 2])
-    study.tell(trial, 0)
+    study.tell(trial, [0, 0] if multiobjective else 0)
 
 
 def test_categorical_distance_func_experimental_warning() -> None: