[tune] Add seed parameter to BOHB (#12160)

python/ray/tune/suggest/bohb.py

@@ -48,6 +48,9 @@ class TuneBOHB(Searcher):
         metric (str): The training result objective value attribute.
         mode (str): One of {min, max}. Determines whether objective is
             minimizing or maximizing the metric attribute.
+        seed (int): Optional random seed to initialize the random number
+            generator. Setting this should lead to identical initial
+            configurations at each run.
 
     Tune automatically converts search spaces to TuneBOHB's format:
 
@@ -100,7 +103,8 @@ class TuneBOHB(Searcher):
                  bohb_config: Optional[Dict] = None,
                  max_concurrent: int = 10,
                  metric: Optional[str] = None,
-                 mode: Optional[str] = None):
+                 mode: Optional[str] = None,
+                 seed: Optional[int] = None):
         from hpbandster.optimizers.config_generators.bohb import BOHB
         assert BOHB is not None, "HpBandSter must be installed!"
         if mode:
@@ -122,6 +126,7 @@ class TuneBOHB(Searcher):
                 space = self.convert_search_space(space)
 
         self._space = space
+        self._seed = seed
 
         super(TuneBOHB, self).__init__(metric=self._metric, mode=mode)
 
@@ -136,6 +141,9 @@ class TuneBOHB(Searcher):
         elif self._mode == "min":
             self._metric_op = 1.
 
+        if self._seed is not None:
+            self._space.seed(self._seed)
+
         bohb_config = self._bohb_config or {}
         self.bohber = BOHB(self._space, **bohb_config)
 

python/ray/tune/tests/test_searchers.py

@@ -35,14 +35,35 @@ class InvalidValuesTest(unittest.TestCase):
     def tearDownClass(cls):
         ray.shutdown()
 
-    def testBayesOpt(self):
-        from ray.tune.suggest.bayesopt import BayesOptSearch
+    def testAx(self):
+        from ray.tune.suggest.ax import AxSearch
+        from ax.service.ax_client import AxClient
 
-        np.random.seed(1234)  # At least one nan, inf, -inf and float
+        converted_config = AxSearch.convert_search_space(self.config)
+        # At least one nan, inf, -inf and float
+        client = AxClient(random_seed=4321)
+        client.create_experiment(
+            parameters=converted_config, objective_name="_metric")
+        searcher = AxSearch(ax_client=client, metric="_metric", mode="max")
 
         out = tune.run(
             _invalid_objective,
-            search_alg=BayesOptSearch(),
+            search_alg=searcher,
+            metric="_metric",
+            mode="max",
+            num_samples=4,
+            reuse_actors=False)
+
+        best_trial = out.best_trial
+        self.assertLessEqual(best_trial.config["report"], 2.0)
+
+    def testBayesOpt(self):
+        from ray.tune.suggest.bayesopt import BayesOptSearch
+
+        out = tune.run(
+            _invalid_objective,
+            # At least one nan, inf, -inf and float
+            search_alg=BayesOptSearch(random_state=1234),
             config=self.config,
             metric="_metric",
             mode="max",
@@ -55,13 +76,10 @@ class InvalidValuesTest(unittest.TestCase):
     def testBOHB(self):
         from ray.tune.suggest.bohb import TuneBOHB
 
-        converted_config = TuneBOHB.convert_search_space(self.config)
-        converted_config.seed(1000)  # At least one nan, inf, -inf and float
-
         out = tune.run(
             _invalid_objective,
-            search_alg=TuneBOHB(
-                space=converted_config, metric="_metric", mode="max"),
+            search_alg=TuneBOHB(seed=1000),
+            config=self.config,
             metric="_metric",
             mode="max",
             num_samples=8,