Remove hypothesis dependency (#3395)

Summary:
Pull Request resolved: #3395

Hypothesis is not commonly used in Ax and is only being used in one model bridge transforms test. Removing it will make it easier to build an environment that can run tests.

With the removal of Hypothesis from this test, we remove Hypothesis from Ax entirely.

Reviewed By: esantorella

Differential Revision: D69858318

fbshipit-source-id: 73771cd389299ebce53789b9f569b9fd4043f8a3

ax/modelbridge/transforms/tests/test_relativize_transform.py

@@ -40,7 +40,6 @@
     get_branin_with_multi_task,
     get_search_space,
 )
-from hypothesis import assume, given, settings, strategies as st
 from pyre_extensions import assert_is_instance, none_throws
 
 
@@ -328,60 +327,51 @@ class BadRelativize(BaseRelativize):
                     modelbridge=None,
                 )
 
-    # pyre-fixme[56]: Pyre was not able to infer the type of argument
-    #  `hypothesis.strategies.floats($parameter$min_value = - 10.000000,
-    #  $parameter$max_value = 10.000000)` to decorator factory `hypothesis.given`.
-    @given(
-        st.floats(min_value=-10.0, max_value=10.0),
-        st.floats(min_value=0, max_value=10.0),
-        st.floats(min_value=-10.0, max_value=10.0),
-        st.floats(min_value=0, max_value=10.0),
-    )
-    @settings(max_examples=1000, deadline=None)
-    def test_transform_status_quos_always_zero(
-        self,
-        sq_mean: float,
-        sq_sem: float,
-        mean: float,
-        sem: float,
-    ) -> None:
-        assume(abs(sq_mean) >= 1e-10)
-        assume(abs(sq_mean) != sq_sem)
-
-        arm_names = ["status_quo", "0_0"]
-        obs_data = [
-            ObservationData(
-                metric_names=["foo"],
-                means=np.array([sq_mean]),
-                covariance=np.array([[sq_sem]]),
-            ),
-            ObservationData(
-                metric_names=["foo"],
-                means=np.array([mean]),
-                covariance=np.array([[sem]]),
-            ),
-        ]
-        obs_features = [
-            ObservationFeatures(parameters={"x": 1}, trial_index=0),
-            ObservationFeatures(parameters={"x": 2}, trial_index=0),
-        ]
-        modelbridge = Mock(
-            status_quo=Mock(
-                data=obs_data[0], features=obs_features[0], arm_name=arm_names[0]
-            ),
-            status_quo_data_by_trial={0: obs_data[0]},
-        )
-        observations = recombine_observations(obs_features, obs_data, arm_names)
-        for relativize_cls in [Relativize, RelativizeWithConstantControl]:
-            transform = relativize_cls(
-                search_space=None,
-                observations=observations,
-                modelbridge=modelbridge,
+    def test_transform_status_quos_always_zero(self) -> None:
+        for _ in range(1000):
+            sq_mean = np.random.uniform(-10.0, 10.0)
+            sq_sem = np.random.uniform(0, 10.0)
+            mean = np.random.uniform(-10.0, 10.0)
+            sem = np.random.uniform(0, 10.0)
+            if abs(sq_mean) < 1e-10 or abs(sq_mean) == sq_sem:
+                continue
+
+            arm_names = ["status_quo", "0_0"]
+            obs_data = [
+                ObservationData(
+                    metric_names=["foo"],
+                    means=np.array([sq_mean]),
+                    covariance=np.array([[sq_sem]]),
+                ),
+                ObservationData(
+                    metric_names=["foo"],
+                    means=np.array([mean]),
+                    covariance=np.array([[sem]]),
+                ),
+            ]
+            obs_features = [
+                ObservationFeatures(parameters={"x": 1}, trial_index=0),
+                ObservationFeatures(parameters={"x": 2}, trial_index=0),
+            ]
+            modelbridge = Mock(
+                status_quo=Mock(
+                    data=obs_data[0], features=obs_features[0], arm_name=arm_names[0]
+                ),
+                status_quo_data_by_trial={0: obs_data[0]},
             )
-            relative_obs = transform.transform_observations(observations)
-            self.assertEqual(relative_obs[0].data.metric_names, ["foo"])
-            self.assertAlmostEqual(relative_obs[0].data.means[0], 0, places=4)
-            self.assertAlmostEqual(relative_obs[0].data.covariance[0][0], 0, places=4)
+            observations = recombine_observations(obs_features, obs_data, arm_names)
+            for relativize_cls in [Relativize, RelativizeWithConstantControl]:
+                transform = relativize_cls(
+                    search_space=None,
+                    observations=observations,
+                    modelbridge=modelbridge,
+                )
+                relative_obs = transform.transform_observations(observations)
+                self.assertEqual(relative_obs[0].data.metric_names, ["foo"])
+                self.assertAlmostEqual(relative_obs[0].data.means[0], 0, places=4)
+                self.assertAlmostEqual(
+                    relative_obs[0].data.covariance[0][0], 0, places=4
+                )
 
     def test_multitask_data(self) -> None:
         experiment = get_branin_with_multi_task()

setup.py

@@ -36,7 +36,6 @@
     "beautifulsoup4",
     "black==24.2.0",
     "flake8",
-    "hypothesis",
     "Jinja2",
     "pyfakefs",
     "pytest>=4.6",