Merge d010419 into db061c1

autokeras/hypermodel/preprocessor.py

@@ -745,7 +745,6 @@ class FeatureEngineering(Preprocessor):
     """
 
     def __init__(self, max_columns=1000, **kwargs):
-
         super().__init__(**kwargs)
         self.input_node = None
         self.max_columns = max_columns

autokeras/meta_model.py

@@ -218,6 +218,10 @@ def assemble(self, input_node):
         self.infer_column_types()
         if input_node.column_types is None:
             input_node.column_types = self.column_types
+        # partial column_types is provided.
+        for key, value in self.column_types.items():
+            if key not in input_node.column_types:
+                input_node.column_types[key] = value
         return hyperblock.StructuredDataBlock()(input_node)
 
 

autokeras/task.py

@@ -176,6 +176,11 @@ def __init__(self, outputs, column_names, column_types, **kwargs):
         inputs = node.StructuredDataInput()
         inputs.column_types = column_types
         inputs.column_names = column_names
+        if column_types:
+            if not all([column_type in ['categorical', 'numerical']
+                        for column_type in column_types.values()]):
+                raise ValueError(
+                    'Column_types should be either "categorical" or "numerical".')
         if column_names and column_types:
             if not all([column_name in column_names
                         for column_name in column_types]):
@@ -185,10 +190,10 @@ def __init__(self, outputs, column_names, column_types, **kwargs):
                          **kwargs)
 
     def fit(self,
-            x=None,  # file path of training data
-            y=None,  # label name
+            x=None,
+            y=None,
             validation_split=0,
-            validation_data=None,  # file path of validataion data
+            validation_data=None,
             **kwargs):
         """
         # Arguments

tests/common.py

@@ -46,6 +46,22 @@
                           'deck_': 'categorical',
                           'embark_town_': 'categorical',
                           'alone_': 'categorical'}
+false_column_types_from_csv = {
+                          'sex_': 'cat',
+                          'age_': 'num',
+                          'n_siblings_spouses_': 'cat',
+                          'parch_': 'categorical',
+                          'fare_': 'numerical',
+                          'class_': 'categorical',
+                          'deck_': 'categorical',
+                          'embark_town_': 'categorical',
+                          'alone_': 'categorical'}
+partial_column_types_from_csv = {
+                          'fare': 'numerical',
+                          'class': 'categorical',
+                          'deck': 'categorical',
+                          'embark_town': 'categorical',
+                          'alone': 'categorical'}
 
 
 def structured_data(num_data=500):

tests/test_task.py

@@ -9,10 +9,11 @@
 from tests.common import column_names_from_numpy
 from tests.common import column_types_from_csv
 from tests.common import column_types_from_numpy
+from tests.common import false_column_types_from_csv
 from tests.common import less_column_names_from_csv
+from tests.common import partial_column_types_from_csv
 from tests.common import structured_data
 
-
 train_file_path = r'tests/resources/titanic/train.csv'
 test_file_path = r'tests/resources/titanic/eval.csv'
 
@@ -88,7 +89,7 @@ def test_structured_data_from_numpy_regressor(tmp_dir):
     x_train = data
     y = np.random.rand(num_data, 1)
     y_train = y
-    clf = ak.StructuredDataRegressor(directory=tmp_dir, max_trials=2)
+    clf = ak.StructuredDataRegressor(directory=tmp_dir, max_trials=1)
     clf.fit(x_train, y_train, epochs=2, validation_data=(x_train, y_train))
 
 
@@ -125,7 +126,7 @@ def test_structured_data_from_numpy_col_type_classifier(tmp_dir):
         clf = ak.StructuredDataClassifier(
             column_types=column_types_from_numpy,
             directory=tmp_dir,
-            max_trials=2)
+            max_trials=1)
         clf.fit(x_train, y_train, epochs=2, validation_data=(x_train, y_train))
     assert str(info.value) == 'Column names must be specified.'
 
@@ -140,7 +141,7 @@ def test_structured_data_from_numpy_col_name_type_classifier(tmp_dir):
         column_names=column_names_from_numpy,
         column_types=column_types_from_numpy,
         directory=tmp_dir,
-        max_trials=2)
+        max_trials=1)
     clf.fit(x_train, y_train, epochs=2, validation_data=(x_train, y_train))
 
 
@@ -151,7 +152,7 @@ def test_structured_data_classifier_transform_new_data(tmp_dir):
     x_train, x_test = data[:num_train], data[num_train:]
     y = np.random.randint(0, 3, num_data)
     y_train, y_test = y[:num_train], y[num_train:]
-    clf = ak.StructuredDataClassifier(directory=tmp_dir, max_trials=2)
+    clf = ak.StructuredDataClassifier(directory=tmp_dir, max_trials=1)
     clf.fit(x_train, y_train, epochs=2, validation_data=(x_test, y_test))
 
 
@@ -205,3 +206,24 @@ def test_structured_data_from_csv_col_type_mismatch_classifier(tmp_dir):
         clf.fit(x=train_file_path, y='survived', epochs=2,
                 validation_data=test_file_path)
     assert str(info.value) == 'Column_names and column_types are mismatched.'
+
+
+def test_structured_data_from_csv_false_col_type_classifier(tmp_dir):
+    with pytest.raises(ValueError) as info:
+        clf = ak.StructuredDataClassifier(
+            column_types=false_column_types_from_csv,
+            directory=tmp_dir,
+            max_trials=1)
+        clf.fit(x=train_file_path, y='survived', epochs=2,
+                validation_data=test_file_path)
+    assert str(info.value) == \
+        'Column_types should be either "categorical" or "numerical".'
+
+
+def test_structured_data_from_csv_partial_col_type_classifier(tmp_dir):
+    clf = ak.StructuredDataClassifier(
+        column_types=partial_column_types_from_csv,
+        directory=tmp_dir,
+        max_trials=1)
+    clf.fit(x=train_file_path, y='survived', epochs=2,
+            validation_data=test_file_path)