Update MTGP transforms to use new MBM_X_trans (#3146)

Summary: Pull Request resolved: #3146 Follow up to D66724547 to propagate the new transforms to MTGP models. Reviewed By: lena-kashtelyan Differential Revision: D66726992 fbshipit-source-id: 1fdeaba5893e79bca909a7a33d14ee4e7a4a046b
facebook · Dec 9, 2024 · be685a7 · be685a7
1 parent 6cf2ae3
commit be685a7
Show file tree

Hide file tree

Showing 4 changed files with 25 additions and 20 deletions.
diff --git a/ax/modelbridge/registry.py b/ax/modelbridge/registry.py
@@ -36,7 +36,6 @@
     ChoiceToNumericChoice,
     OrderedChoiceToIntegerRange,
 )
-from ax.modelbridge.transforms.convert_metric_names import ConvertMetricNames
 from ax.modelbridge.transforms.derelativize import Derelativize
 from ax.modelbridge.transforms.fill_missing_parameters import FillMissingParameters
 from ax.modelbridge.transforms.int_range_to_choice import IntRangeToChoice
@@ -131,15 +130,6 @@
 # call `list.__add__` but got `List[Type[SearchSpaceToChoice]]`.
 TS_trans: list[type[Transform]] = Y_trans + [SearchSpaceToChoice]
 
-# Multi-type MTGP transforms
-MT_MTGP_trans: list[type[Transform]] = Cont_X_trans + [
-    Derelativize,
-    ConvertMetricNames,
-    TrialAsTask,
-    StratifiedStandardizeY,
-    TaskChoiceToIntTaskChoice,
-]
-
 # Single-type MTGP transforms
 ST_MTGP_trans: list[type[Transform]] = Cont_X_trans + [
     Derelativize,
@@ -148,9 +138,9 @@
     TaskChoiceToIntTaskChoice,
 ]
 
-# Single-type MTGP transforms
-Specified_Task_ST_MTGP_trans: list[type[Transform]] = Cont_X_trans + [
+MBM_MTGP_trans: list[type[Transform]] = MBM_X_trans + [
     Derelativize,
+    TrialAsTask,
     StratifiedStandardizeY,
     TaskChoiceToIntTaskChoice,
 ]
@@ -218,7 +208,7 @@ class ModelSetup(NamedTuple):
     "ST_MTGP": ModelSetup(
         bridge_class=TorchModelBridge,
         model_class=ModularBoTorchModel,
-        transforms=ST_MTGP_trans,
+        transforms=MBM_MTGP_trans,
         standard_bridge_kwargs=STANDARD_TORCH_BRIDGE_KWARGS,
     ),
     "BO_MIXED": ModelSetup(
@@ -241,7 +231,7 @@ class ModelSetup(NamedTuple):
     "SAAS_MTGP": ModelSetup(
         bridge_class=TorchModelBridge,
         model_class=ModularBoTorchModel,
-        transforms=ST_MTGP_trans,
+        transforms=MBM_MTGP_trans,
         default_model_kwargs={
             "surrogate_spec": SurrogateSpec(
                 botorch_model_class=SaasFullyBayesianMultiTaskGP

diff --git a/ax/modelbridge/tests/test_generation_strategy.py b/ax/modelbridge/tests/test_generation_strategy.py
@@ -43,9 +43,9 @@
 from ax.modelbridge.registry import (
     _extract_model_state_after_gen,
     Cont_X_trans,
+    MBM_MTGP_trans,
     MODEL_KEY_TO_MODEL_SETUP,
     Models,
-    ST_MTGP_trans,
 )
 from ax.modelbridge.torch import TorchModelBridge
 from ax.modelbridge.transition_criterion import (
@@ -1106,7 +1106,7 @@ def test_gen_for_multiple_trials_with_multiple_models_with_fixed_features(
                     model_kwargs={
                         # this will cause an error if the model
                         # doesn't get fixed features
-                        "transforms": ST_MTGP_trans,
+                        "transforms": MBM_MTGP_trans,
                         **self.step_model_kwargs,
                     },
                     num_trials=1,

diff --git a/ax/service/tests/scheduler_test_utils.py b/ax/service/tests/scheduler_test_utils.py
@@ -44,7 +44,7 @@
 from ax.modelbridge.cross_validation import compute_model_fit_metrics_from_modelbridge
 from ax.modelbridge.dispatch_utils import choose_generation_strategy
 from ax.modelbridge.generation_strategy import GenerationStep, GenerationStrategy
-from ax.modelbridge.registry import Models, ST_MTGP_trans
+from ax.modelbridge.registry import MBM_MTGP_trans, Models
 from ax.runners.single_running_trial_mixin import SingleRunningTrialMixin
 from ax.runners.synthetic import SyntheticRunner
 from ax.service.scheduler import (
@@ -2391,7 +2391,7 @@ def test_it_works_with_multitask_models(
                         model_kwargs={
                             # this will cause and error if the model
                             # doesn't get fixed features
-                            "transforms": ST_MTGP_trans,
+                            "transforms": MBM_MTGP_trans,
                             "transform_configs": {
                                 "TrialAsTask": {
                                     "trial_level_map": {

diff --git a/tutorials/multi_task.ipynb b/tutorials/multi_task.ipynb
@@ -56,15 +56,30 @@
     "from ax.core.search_space import SearchSpace\n",
     "from ax.metrics.hartmann6 import Hartmann6Metric\n",
     "from ax.modelbridge.factory import get_sobol\n",
-    "from ax.modelbridge.registry import Models, MT_MTGP_trans, ST_MTGP_trans\n",
+    "from ax.modelbridge.registry import Models, MBM_X_trans, ST_MTGP_trans\n",
     "from ax.modelbridge.torch import TorchModelBridge\n",
     "from ax.modelbridge.transforms.convert_metric_names import tconfig_from_mt_experiment\n",
+    "from ax.modelbridge.transforms.derelativize import Derelativize\n",
+    "from ax.modelbridge.transforms.convert_metric_names import ConvertMetricNames\n",
+    "from ax.modelbridge.transforms.trial_as_task import TrialAsTask\n",
+    "from ax.modelbridge.transforms.stratified_standardize_y import StratifiedStandardizeY\n",
+    "from ax.modelbridge.transforms.task_encode import TaskChoiceToIntTaskChoice\n",
     "from ax.plot.diagnostic import interact_batch_comparison\n",
     "from ax.runners.synthetic import SyntheticRunner\n",
     "from ax.utils.common.typeutils import checked_cast\n",
     "from ax.utils.notebook.plotting import init_notebook_plotting, render\n",
     "\n",
-    "init_notebook_plotting()"
+    "init_notebook_plotting()\n",
+    "\n",
+    "# Transforms for pre-processing the data from a multi-type experiment to \n",
+    "# construct a multi-task GP model.\n",
+    "MT_MTGP_trans = MBM_X_trans + [\n",
+    "    Derelativize,\n",
+    "    ConvertMetricNames,\n",
+    "    TrialAsTask,\n",
+    "    StratifiedStandardizeY,\n",
+    "    TaskChoiceToIntTaskChoice,\n",
+    "]"
    ]
   },
   {