add tests for quantized training with categorical features

tests/python_package_test/test_engine.py

@@ -309,7 +309,8 @@ def test_missing_value_handle_none():
     assert evals_result['valid_0']['auc'][-1] == pytest.approx(ret)
 
 
-def test_categorical_handle():
+@pytest.mark.parametrize('use_quantized_grad', [True, False])
+def test_categorical_handle(use_quantized_grad):
     x = [0, 1, 2, 3, 4, 5, 6, 7]
     y = [0, 1, 0, 1, 0, 1, 0, 1]
 
@@ -332,7 +333,8 @@ def test_categorical_handle():
         'cat_l2': 0,
         'max_cat_to_onehot': 1,
         'zero_as_missing': True,
-        'categorical_column': 0
+        'categorical_column': 0,
+        'use_quantized_grad': use_quantized_grad,
     }
     evals_result = {}
     gbm = lgb.train(
@@ -349,7 +351,8 @@ def test_categorical_handle():
     assert evals_result['valid_0']['auc'][-1] == pytest.approx(ret)
 
 
-def test_categorical_handle_na():
+@pytest.mark.parametrize('use_quantized_grad', [True, False])
+def test_categorical_handle_na(use_quantized_grad):
     x = [0, np.nan, 0, np.nan, 0, np.nan]
     y = [0, 1, 0, 1, 0, 1]
 
@@ -372,7 +375,8 @@ def test_categorical_handle_na():
         'cat_l2': 0,
         'max_cat_to_onehot': 1,
         'zero_as_missing': False,
-        'categorical_column': 0
+        'categorical_column': 0,
+        'use_quantized_grad': use_quantized_grad,
     }
     evals_result = {}
     gbm = lgb.train(
@@ -389,7 +393,8 @@ def test_categorical_handle_na():
     assert evals_result['valid_0']['auc'][-1] == pytest.approx(ret)
 
 
-def test_categorical_non_zero_inputs():
+@pytest.mark.parametrize('use_quantized_grad', [True, False])
+def test_categorical_non_zero_inputs(use_quantized_grad):
     x = [1, 1, 1, 1, 1, 1, 2, 2]
     y = [1, 1, 1, 1, 1, 1, 0, 0]
 
@@ -412,7 +417,8 @@ def test_categorical_non_zero_inputs():
         'cat_l2': 0,
         'max_cat_to_onehot': 1,
         'zero_as_missing': False,
-        'categorical_column': 0
+        'categorical_column': 0,
+        'use_quantized_grad': use_quantized_grad,
     }
     evals_result = {}
     gbm = lgb.train(