Make models enum non-optional for get_model_from_generator_run

Summary: This being optional was creating a bug-prone codepath due to our use of the method on generator runs using internal only models. This change to the models enum being required cascaded through a fairly large portion of our code, stemming from our handling of best_point functions, but I think it will be worth it and not as disruptive as one may think.

Reviewed By: lena-kashtelyan

Differential Revision: D33238024

fbshipit-source-id: ce3bd7089ad208f19c3b88d7b5d22483e2b59986

Author: mpolson64

torchx/runtime/hpo/test/ax_test.py

@@ -22,6 +22,7 @@
     SearchSpace,
 )
 from ax.modelbridge.dispatch_utils import choose_generation_strategy
+from ax.modelbridge.registry import Models
 from ax.service.scheduler import SchedulerOptions
 from ax.service.utils.best_point import get_best_parameters
 from ax.service.utils.report_utils import exp_to_df
@@ -104,7 +105,9 @@ def test_run_experiment_locally(self) -> None:
 
         # AppMetrics always returns trial index; hence the best
         # experiment for min objective will be the params for trial 0
-        best_param, _ = none_throws(get_best_parameters(experiment))
+        best_param, _ = none_throws(
+            get_best_parameters(experiment=experiment, models_enum=Models)
+        )
         # nothing to assert, just make sure experiment runs
 
     def test_run_experiment_locally_in_batches(self) -> None:
@@ -150,7 +153,9 @@ def test_run_experiment_locally_in_batches(self) -> None:
 
         # AppMetrics always returns trial index; hence the best
         # experiment for min objective will be the params for trial 0
-        best_param, _ = none_throws(get_best_parameters(experiment))
+        best_param, _ = none_throws(
+            get_best_parameters(experiment=experiment, models_enum=Models)
+        )
         # nothing to assert, just make sure experiment runs
 
     def test_runner_no_batch_trials(self) -> None: