add flows by default to flows generated by sklearn converter

Author: mfeurer

openml/flows/functions.py

@@ -208,7 +208,15 @@ def assert_flows_equal(flow1, flow2, ignore_parameters_on_older_children=None,
                 elif ignore_parameters:
                     continue
 
-            if attr1 != attr2:
+            if key == 'tags':
+                if set(attr1) != set(attr2):
+                    raise ValueError(
+                        "Flow %s: values for attribute '%s' differ: "
+                        "'%s' vs '%s'." %
+                        (str(flow1.name), str(key), str(set(attr1)),
+                         str(set(attr2))))
+
+            elif attr1 != attr2:
                 raise ValueError("Flow %s: values for attribute '%s' differ: "
                                  "'%s' vs '%s'." %
                                  (str(flow1.name), str(key), str(attr1), str(attr2)))

openml/flows/sklearn_converter.py

@@ -208,7 +208,15 @@ def _serialize_model(model):
                       parameters=parameters,
                       parameters_meta_info=parameters_meta_info,
                       external_version=external_version,
-                      tags=[],
+                      tags=['openml-python', 'sklearn', 'scikit-learn',
+                            'python',
+                            _format_external_version('sklearn',
+                                                     sklearn.__version__).replace('==', '_'),
+                            # TODO: add more tags based on the scikit-learn
+                            # module a flow is in? For example automatically
+                            # annotate a class of sklearn.svm.SVC() with the
+                            # tag svm?
+                            ],
                       language='English',
                       # TODO fill in dependencies!
                       dependencies=dependencies)

requirements.txt

@@ -7,3 +7,4 @@ nose
 requests
 scikit-learn>=0.18
 nbformat
+dateutil

tests/test_flows/test_flow.py

@@ -276,7 +276,14 @@ def test_sklearn_to_upload_to_flow(self):
             estimator=model, param_distributions=parameter_grid, cv=cv)
         rs.fit(X, y)
         flow = openml.flows.sklearn_to_flow(rs)
-        flow.tags.extend(['openml-python', 'unittest'])
+        # Tags may be sorted in any order (by the server). Just using one tag
+        # makes sure that the xml comparison does not fail because of that.
+        subflows = [flow]
+        while len(subflows) > 0:
+            f = subflows.pop()
+            f.tags = []
+            subflows.extend(list(f.components.values()))
+
         flow, sentinel = self._add_sentinel_to_flow_name(flow, None)
 
         flow.publish()
@@ -317,8 +324,6 @@ def test_sklearn_to_upload_to_flow(self):
                         % sentinel
 
         self.assertEqual(new_flow.name, fixture_name)
-        self.assertTrue('openml-python' in new_flow.tags)
-        self.assertTrue('unittest' in new_flow.tags)
         new_flow.model.fit(X, y)
 
     def test_extract_tags(self):