Make fitting compatible with scipy 1.15 optimization changes (#2667)

Summary:
Resolves #2666 by updating the regexp that is used to filter out the optimization warning emitted when the maximium number of iterations is reached.

NOTE: This kind of regexp filtering is kind of brittle, but it's not necessarily obvious how to do this differently if scipy doesn't return these in a more structured form.

Pull Request resolved: #2667

Reviewed By: saitcakmak

Differential Revision: D67816653

Pulled By: Balandat

fbshipit-source-id: aa6f83298e50e4e1ebedde7893888ecd82c4bae1

botorch/optim/core.py

@@ -34,7 +34,9 @@
 
 
 _LBFGSB_MAXITER_MAXFUN_REGEX = re.compile(  # regex for maxiter and maxfun messages
-    "TOTAL NO. of (ITERATIONS REACHED LIMIT|f AND g EVALUATIONS EXCEEDS LIMIT)"
+    # Note that the messages changed with scipy 1.15, hence the different matching here.
+    "TOTAL NO. (of|OF) "
+    + "(ITERATIONS REACHED LIMIT|(f AND g|F,G) EVALUATIONS EXCEEDS LIMIT)"
 )
 
 

requirements.txt

@@ -4,5 +4,5 @@ gpytorch==1.13
 linear_operator==0.5.3
 torch>=2.0.1
 pyro-ppl>=1.8.4
-scipy<1.15
+scipy
 multipledispatch

test/generation/test_gen.py

@@ -5,6 +5,7 @@
 # LICENSE file in the root directory of this source tree.
 
 import math
+import re
 import warnings
 from unittest import mock
 
@@ -225,13 +226,14 @@ def test_gen_candidates_scipy_with_fixed_features_inequality_constraints(self):
     def test_gen_candidates_scipy_warns_opt_failure(self):
         with warnings.catch_warnings(record=True) as ws:
             self.test_gen_candidates(options={"maxls": 1})
-        expected_msg = (
+        expected_msg = re.compile(
+            # The message changed with scipy 1.15, hence the different matching here.
             "Optimization failed within `scipy.optimize.minimize` with status 2"
-            " and message ABNORMAL_TERMINATION_IN_LNSRCH."
+            " and message ABNORMAL(|_TERMINATION_IN_LNSRCH)."
         )
         expected_warning_raised = any(
             issubclass(w.category, OptimizationWarning)
-            and expected_msg in str(w.message)
+            and expected_msg.search(str(w.message))
             for w in ws
         )
         self.assertTrue(expected_warning_raised)

test/optim/test_core.py

@@ -135,11 +135,20 @@ def _callback(parameters, result, out) -> None:
     def test_post_processing(self):
         closure = next(iter(self.closures.values()))
         wrapper = NdarrayOptimizationClosure(closure, closure.parameters)
+
+        # Scipy changed return values and messages in v1.15, so we check both
+        # old and new versions here.
+        status_msgs = [
+            # scipy >=1.15
+            (OptimizationStatus.FAILURE, "ABNORMAL_TERMINATION_IN_LNSRCH"),
+            (OptimizationStatus.STOPPED, "TOTAL NO. of ITERATIONS REACHED LIMIT"),
+            # scipy <1.15
+            (OptimizationStatus.FAILURE, "ABNORMAL "),
+            (OptimizationStatus.STOPPED, "TOTAL NO. OF ITERATIONS REACHED LIMIT"),
+        ]
+
         with patch.object(core, "minimize_with_timeout") as mock_minimize_with_timeout:
-            for status, msg in (
-                (OptimizationStatus.FAILURE, b"ABNORMAL_TERMINATION_IN_LNSRCH"),
-                (OptimizationStatus.STOPPED, "TOTAL NO. of ITERATIONS REACHED LIMIT"),
-            ):
+            for status, msg in status_msgs:
                 mock_minimize_with_timeout.return_value = OptimizeResult(
                     x=wrapper.state,
                     fun=1.0,

test/optim/test_fit.py

@@ -5,6 +5,7 @@
 # LICENSE file in the root directory of this source tree.
 
 import math
+import re
 from unittest.mock import MagicMock, patch
 from warnings import catch_warnings
 
@@ -20,6 +21,11 @@
 from gpytorch.mlls.exact_marginal_log_likelihood import ExactMarginalLogLikelihood
 from scipy.optimize import OptimizeResult
 
+MAX_ITER_MSG_REGEX = re.compile(
+    # Note that the message changed with scipy 1.15, hence the different matching here.
+    "TOTAL NO. (of|OF) ITERATIONS REACHED LIMIT"
+)
+
 
 class TestFitGPyTorchMLLScipy(BotorchTestCase):
     def setUp(self, suppress_input_warnings: bool = True) -> None:
@@ -63,15 +69,18 @@ def _test_fit_gpytorch_mll_scipy(self, mll):
             )
 
             # Test maxiter warning message
-            self.assertTrue(any("TOTAL NO. of" in str(w.message) for w in ws))
+
+            self.assertTrue(any(MAX_ITER_MSG_REGEX.search(str(w.message)) for w in ws))
             self.assertTrue(
                 any(issubclass(w.category, OptimizationWarning) for w in ws)
             )
 
             # Test iteration tracking
             self.assertIsInstance(result, OptimizationResult)
             self.assertLessEqual(result.step, options["maxiter"])
-            self.assertEqual(sum(1 for w in ws if "TOTAL NO. of" in str(w.message)), 1)
+            self.assertEqual(
+                sum(1 for w in ws if MAX_ITER_MSG_REGEX.search(str(w.message))), 1
+            )
 
         # Test that user provided bounds are respected
         with self.subTest("bounds"), module_rollback_ctx(mll, checkpoint=ckpt):

test/optim/test_optimize.py

@@ -5,6 +5,7 @@
 # LICENSE file in the root directory of this source tree.
 
 import itertools
+import re
 import warnings
 from functools import partial
 from itertools import product
@@ -724,19 +725,20 @@ def test_optimize_acqf_warns_on_opt_failure(self):
                 raw_samples=raw_samples,
                 batch_initial_conditions=initial_conditions,
             )
-        message = (
-            "Optimization failed in `gen_candidates_scipy` with the following "
-            "warning(s):\n[OptimizationWarning('Optimization failed within "
-            "`scipy.optimize.minimize` with status 2 and message "
-            "ABNORMAL_TERMINATION_IN_LNSRCH.')]\nBecause you specified "
-            "`batch_initial_conditions` larger than required `num_restarts`, "
-            "optimization will not be retried with new initial conditions and "
-            "will proceed with the current solution. Suggested remediation: "
-            "Try again with different `batch_initial_conditions`, don't provide "
-            "`batch_initial_conditions`, or increase `num_restarts`."
+        message_regex = re.compile(
+            r"Optimization failed in `gen_candidates_scipy` with the following "
+            r"warning\(s\):\n\[OptimizationWarning\('Optimization failed within "
+            r"`scipy.optimize.minimize` with status 2 and message "
+            r"ABNORMAL(: |_TERMINATION_IN_LNSRCH).'\)]\nBecause you specified "
+            r"`batch_initial_conditions` larger than required `num_restarts`, "
+            r"optimization will not be retried with new initial conditions and "
+            r"will proceed with the current solution. Suggested remediation: "
+            r"Try again with different `batch_initial_conditions`, don't provide "
+            r"`batch_initial_conditions`, or increase `num_restarts`."
         )
         expected_warning_raised = any(
-            issubclass(w.category, RuntimeWarning) and message in str(w.message)
+            issubclass(w.category, RuntimeWarning)
+            and message_regex.search(str(w.message))
             for w in ws
         )
         self.assertTrue(expected_warning_raised)
@@ -774,14 +776,16 @@ def test_optimize_acqf_successfully_restarts_on_opt_failure(self):
                 # more likely
                 options={"maxls": 2},
             )
-        message = (
-            "Optimization failed in `gen_candidates_scipy` with the following "
-            "warning(s):\n[OptimizationWarning('Optimization failed within "
-            "`scipy.optimize.minimize` with status 2 and message ABNORMAL_TERMINATION"
-            "_IN_LNSRCH.')]\nTrying again with a new set of initial conditions."
+        message_regex = re.compile(
+            r"Optimization failed in `gen_candidates_scipy` with the following "
+            r"warning\(s\):\n\[OptimizationWarning\('Optimization failed within "
+            r"`scipy.optimize.minimize` with status 2 and message ABNORMAL(: |"
+            r"_TERMINATION_IN_LNSRCH).'\)\]\nTrying again with a new set of "
+            r"initial conditions."
         )
         expected_warning_raised = any(
-            issubclass(w.category, RuntimeWarning) and message in str(w.message)
+            issubclass(w.category, RuntimeWarning)
+            and message_regex.search(str(w.message))
             for w in ws
         )
         self.assertTrue(expected_warning_raised)
@@ -803,7 +807,8 @@ def test_optimize_acqf_successfully_restarts_on_opt_failure(self):
                 retry_on_optimization_warning=False,
             )
         expected_warning_raised = any(
-            issubclass(w.category, RuntimeWarning) and message in str(w.message)
+            issubclass(w.category, RuntimeWarning)
+            and message_regex.search(str(w.message))
             for w in ws
         )
         self.assertFalse(expected_warning_raised)
@@ -840,19 +845,21 @@ def test_optimize_acqf_warns_on_second_opt_failure(self):
                 options={"maxls": 2},
             )
 
-        message_1 = (
-            "Optimization failed in `gen_candidates_scipy` with the following "
-            "warning(s):\n[OptimizationWarning('Optimization failed within "
-            "`scipy.optimize.minimize` with status 2 and message ABNORMAL_TERMINATION"
-            "_IN_LNSRCH.')]\nTrying again with a new set of initial conditions."
+        message_1_regex = re.compile(
+            r"Optimization failed in `gen_candidates_scipy` with the following "
+            r"warning\(s\):\n\[OptimizationWarning\('Optimization failed within "
+            r"`scipy.optimize.minimize` with status 2 and message ABNORMAL(: |"
+            r"_TERMINATION_IN_LNSRCH).'\)\]\nTrying again with a new set of "
+            r"initial conditions."
         )
 
         message_2 = (
             "Optimization failed on the second try, after generating a new set "
             "of initial conditions."
         )
         first_expected_warning_raised = any(
-            issubclass(w.category, RuntimeWarning) and message_1 in str(w.message)
+            issubclass(w.category, RuntimeWarning)
+            and message_1_regex.search(str(w.message))
             for w in ws
         )
         second_expected_warning_raised = any(

test/test_fit.py

@@ -5,6 +5,7 @@
 # LICENSE file in the root directory of this source tree.
 
 import math
+import re
 from collections.abc import Callable, Iterable
 from contextlib import ExitStack, nullcontext
 from copy import deepcopy
@@ -30,7 +31,10 @@
 from gpytorch.mlls import ExactMarginalLogLikelihood, VariationalELBO
 from linear_operator.utils.errors import NotPSDError
 
-MAX_ITER_MSG = "TOTAL NO. of ITERATIONS REACHED LIMIT"
+MAX_ITER_MSG_REGEX = re.compile(
+    # Note that the message changed with scipy 1.15, hence the different matching here.
+    "TOTAL NO. (of|OF) ITERATIONS REACHED LIMIT"
+)
 
 
 class MockOptimizer:
@@ -215,7 +219,12 @@ def _test_warnings(self, mll, ckpt):
         optimizer = MockOptimizer(randomize_requires_grad=False)
         optimizer.warnings = [
             WarningMessage("test_runtime_warning", RuntimeWarning, __file__, 0),
-            WarningMessage(MAX_ITER_MSG, OptimizationWarning, __file__, 0),
+            WarningMessage(
+                "STOP: TOTAL NO. OF ITERATIONS REACHED LIMIT",
+                OptimizationWarning,
+                __file__,
+                0,
+            ),
             WarningMessage(
                 "Optimization timed out after X", OptimizationWarning, __file__, 0
             ),
@@ -260,7 +269,9 @@ def _test_warnings(self, mll, ckpt):
                     {str(w.message) for w in rethrown + unresolved},
                 )
                 if logs:  # test that default filter logs certain warnings
-                    self.assertTrue(any(MAX_ITER_MSG in log for log in logs.output))
+                    self.assertTrue(
+                        any(MAX_ITER_MSG_REGEX.search(log) for log in logs.output)
+                    )
 
         # Test default of retrying upon encountering an uncaught OptimizationWarning
         optimizer.warnings.append(

test/test_utils/test_mock.py

@@ -5,6 +5,7 @@
 # LICENSE file in the root directory of this source tree.
 
 
+import re
 import warnings
 from unittest.mock import patch
 
@@ -26,6 +27,12 @@
 from botorch.utils.testing import BotorchTestCase, MockAcquisitionFunction
 
 
+MAX_ITER_MSG = re.compile(
+    # Note that the message changed with scipy 1.15, hence the different matching here.
+    "TOTAL NO. (of|OF) ITERATIONS REACHED LIMIT"
+)
+
+
 class SinAcqusitionFunction(MockAcquisitionFunction):
     """Simple acquisition function with known numerical properties."""
 
@@ -56,9 +63,7 @@ def closure():
 
             with mock_optimize_context_manager():
                 result = scipy_minimize(closure=closure, parameters={"x": x})
-            self.assertEqual(
-                result.message, "STOP: TOTAL NO. of ITERATIONS REACHED LIMIT"
-            )
+            self.assertTrue(MAX_ITER_MSG.search(result.message))
 
         with self.subTest("optimize_acqf"):
             with mock_optimize_context_manager():