fea: naive fix to the init with inequalities

facebook · Oct 24, 2024 · c3793bf · c3793bf
1 parent f66f154
commit c3793bf
Show file tree

Hide file tree

Showing 2 changed files with 18 additions and 8 deletions.
diff --git a/ax/models/torch/botorch_modular/sebo.py b/ax/models/torch/botorch_modular/sebo.py
@@ -298,14 +298,23 @@ def _optimize_with_homotopy(
                 )
             ],
         )
-        batch_initial_conditions = get_batch_initial_conditions(
-            acq_function=self.acqf,
-            raw_samples=optimizer_options_with_defaults["raw_samples"],
-            X_pareto=self.acqf.X_baseline,
-            target_point=self.target_point,
-            bounds=bounds,
-            num_restarts=optimizer_options_with_defaults["num_restarts"],
-        )
+        if inequality_constraints is None:
+            batch_initial_conditions = get_batch_initial_conditions(
+                acq_function=self.acqf,
+                raw_samples=optimizer_options_with_defaults["raw_samples"],
+                X_pareto=self.acqf.X_baseline,
+                target_point=self.target_point,
+                bounds=bounds,
+                num_restarts=optimizer_options_with_defaults["num_restarts"],
+            )
+        else:
+            warnings.warn(
+                "Adopting Botorch's default for sampling initial conditions as "
+                "inequality constraints are not supported for custom SEBO "
+                "initialization based on pareto front perturbation."
+            )
+            batch_initial_conditions = None
+
         candidates, expected_acquisition_value = optimize_acqf_homotopy(
             q=n,
             acq_function=self.acqf,

diff --git a/ax/models/torch/tests/test_sebo.py b/ax/models/torch/tests/test_sebo.py
@@ -310,6 +310,7 @@ def test_optimize_l0(self, mock_optimize_acqf_homotopy: Mock) -> None:
         self.assertEqual(kwargs["q"], 2)
         self.assertEqual(kwargs["inequality_constraints"], self.inequality_constraints)
         self.assertEqual(kwargs["post_processing_func"], self.rounding_func)
+        self.assertEqual(kwargs["batch_initial_conditions"], None)
         self.assertEqual(kwargs["num_restarts"], self.optimizer_options["num_restarts"])
         self.assertEqual(kwargs["raw_samples"], self.optimizer_options["raw_samples"])