support latest xgboost version (#599)

* support latest xgboost version

* Update test_classification.py

* Update 

Exists problems when installing xgb1.6.1 in py3.6

* cleanup

* xgboost version

* remove time_budget_s in test

* remove redundancy

* stop support of python 3.6

Co-authored-by: zsk <shaokunzhang529@gmail.com>
Co-authored-by: Qingyun Wu <qingyun.wu@psu.edu>

test/automl/test_classification.py

@@ -85,6 +85,22 @@ class TestClassification(unittest.TestCase):
         )
         y = pd.Series([0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1])
 
+        automl = AutoML()
+        automl_settings = {
+            "time_budget": 3,
+            "task": "classification",
+            "n_jobs": 1,
+            "estimator_list": ["xgboost", "catboost", "kneighbor"],
+            "eval_method": "cv",
+            "n_splits": 3,
+            "metric": "accuracy",
+            "log_training_metric": True,
+            # "verbose": 4,
+            "ensemble": True,
+        }
+        automl.fit(X, y, **automl_settings)
+        del automl
+
         automl = AutoML()
         automl_settings = {
             "time_budget": 6,
@@ -99,6 +115,7 @@ class TestClassification(unittest.TestCase):
             "ensemble": True,
         }
         automl.fit(X, y, **automl_settings)
+        del automl
 
         automl = AutoML()
         try:
@@ -121,21 +138,7 @@ class TestClassification(unittest.TestCase):
             "n_concurrent_trials": n_concurrent_trials,
         }
         automl.fit(X, y, **automl_settings)
-
-        automl = AutoML()
-        automl_settings = {
-            "time_budget": 3,
-            "task": "classification",
-            "n_jobs": 1,
-            "estimator_list": ["xgboost", "catboost", "kneighbor"],
-            "eval_method": "cv",
-            "n_splits": 3,
-            "metric": "accuracy",
-            "log_training_metric": True,
-            # "verbose": 4,
-            "ensemble": True,
-        }
-        automl.fit(X, y, **automl_settings)
+        del automl
 
         automl = AutoML()
         automl_settings = {
@@ -151,6 +154,7 @@ class TestClassification(unittest.TestCase):
             "ensemble": True,
         }
         automl.fit(X, y, **automl_settings)
+        del automl
 
     def test_binary(self):
         automl_experiment = AutoML()
@@ -208,7 +212,7 @@ class TestClassification(unittest.TestCase):
         _ = automl_experiment.predict(fake_df)
 
     def test_sparse_matrix_xgboost(self):
-        automl_experiment = AutoML()
+        automl = AutoML()
         automl_settings = {
             "time_budget": 3,
             "metric": "ap",
@@ -223,15 +227,28 @@ class TestClassification(unittest.TestCase):
         import xgboost as xgb
 
         callback = xgb.callback.TrainingCallback()
-        automl_experiment.fit(
+        automl.fit(
             X_train=X_train, y_train=y_train, callbacks=[callback], **automl_settings
         )
-        print(automl_experiment.predict(X_train))
-        print(automl_experiment.model)
-        print(automl_experiment.config_history)
-        print(automl_experiment.best_model_for_estimator("xgboost"))
-        print(automl_experiment.best_iteration)
-        print(automl_experiment.best_estimator)
+        print(automl.predict(X_train))
+        print(automl.model)
+        print(automl.config_history)
+        print(automl.best_model_for_estimator("xgboost"))
+        print(automl.best_iteration)
+        print(automl.best_estimator)
+
+        # test an old version of xgboost
+        import subprocess
+        import sys
+
+        subprocess.check_call(
+            [sys.executable, "-m", "pip", "install", "xgboost==1.3.3", "--user"]
+        )
+        automl = AutoML()
+        automl.fit(X_train=X_train, y_train=y_train, **automl_settings)
+        subprocess.check_call(
+            [sys.executable, "-m", "pip", "install", "-U", "xgboost", "--user"]
+        )
 
     def test_ray_classification(self):
         X, y = load_breast_cancer(return_X_y=True)
@@ -353,5 +370,5 @@ class TestClassification(unittest.TestCase):
 
 
 if __name__ == "__main__":
-    unittest.main()
     test = TestClassification()
+    test.test_preprocess()

test/tune/test_reproducibility.py

@@ -24,9 +24,7 @@ def _easy_objective(use_raytune, config):
             return
 
 
-def test_tune(
-    smoke_test=True, externally_setup_searcher=False, use_ray=False, use_raytune=False
-):
+def test_tune(externally_setup_searcher=False, use_ray=False, use_raytune=False):
     from flaml import tune
     from flaml.searcher.blendsearch import BlendSearch
 
@@ -95,7 +93,7 @@ def test_tune(
         metric="mean_loss",
         mode="min",
         num_samples=10,
-        time_budget_s=5,
+        # time_budget_s=5,
         use_ray=use_ray,
         config=search_space,
     )
@@ -107,14 +105,14 @@ def test_tune(
 
 
 def test_reproducibility():
-    best_config_1 = test_tune(smoke_test=True)
-    best_config_2 = test_tune(smoke_test=True)
+    best_config_1 = test_tune()
+    best_config_2 = test_tune()
     print(best_config_1)
     print(best_config_2)
     assert best_config_1 == best_config_2, "flaml.tune not reproducible"
 
-    best_config_1 = test_tune(smoke_test=True, externally_setup_searcher=True)
-    best_config_2 = test_tune(smoke_test=True, externally_setup_searcher=True)
+    best_config_1 = test_tune(externally_setup_searcher=True)
+    best_config_2 = test_tune(externally_setup_searcher=True)
     print(best_config_1)
     print(best_config_2)
     assert (