also moved the function that decides whether a model is traceable to the sklearn connector

Author: janvanrijn

openml/flows/sklearn_converter.py

@@ -550,3 +550,27 @@ def _deserialize_cross_validator(value, **kwargs):
 
 def _format_external_version(model_package_name, model_package_version_number):
     return '%s==%s' % (model_package_name, model_package_version_number)
+
+
+def get_traceble_model(model):
+    '''
+    Returns whether the model can produce an openml trace. If yes,
+    return that model. Returns false otherwise.
+    Clause holds true for instances of BaseSearchCV and Pipelines that
+    contain exactly one BaseSearchCV (no use to search deeper in the
+    Tree, maybe later)
+    '''
+    if isinstance(model, sklearn.model_selection._search.BaseSearchCV):
+        return True
+    count = 0
+    returnValue = None
+    if isinstance(model, sklearn.pipeline.Pipeline):
+        for step in model.steps:
+            if isinstance(step[1], sklearn.model_selection._search.BaseSearchCV):
+                returnValue = step[1]
+                count += 1
+    if count == 1:
+        return returnValue
+    else:
+        return False
+

openml/runs/functions.py

@@ -5,11 +5,10 @@
 import numpy as np
 import warnings
 import sklearn
-from sklearn.model_selection._search import BaseSearchCV
 
 from build.lib.openml.exceptions import PyOpenMLError
 from .. import config
-from ..flows import sklearn_to_flow, get_flow
+from ..flows import sklearn_to_flow, get_flow, get_traceble_model
 from ..setups import setup_exists
 from ..exceptions import OpenMLCacheException, OpenMLServerException
 from ..util import URLError
@@ -158,7 +157,7 @@ def _run_task_get_arffcontent(model, task, class_labels):
             try:
                 model_fold.fit(trainX, trainY)
 
-                if isinstance(model_fold, BaseSearchCV):
+                if get_traceble_model(model_fold):
                     arff_tracecontent.extend(_extract_arfftrace(model_fold, rep_no, fold_no))
                     model_classes = model_fold.best_estimator_.classes_
                 else:
@@ -179,12 +178,12 @@ def _run_task_get_arffcontent(model, task, class_labels):
             fold_no = fold_no + 1
         rep_no = rep_no + 1
 
-    if not isinstance(model, BaseSearchCV):
-        arff_tracecontent = None
-        arff_trace_attributes = None
-    else:
+    if get_traceble_model(model):
         # arff_tracecontent is already set
         arff_trace_attributes = _extract_arfftrace_attributes(model_fold)
+    else:
+        arff_tracecontent = None
+        arff_trace_attributes = None
 
     return arff_datacontent, arff_tracecontent, arff_trace_attributes
 

openml/runs/run.py

@@ -5,7 +5,6 @@
 import arff
 import xmltodict
 from sklearn.base import BaseEstimator
-from sklearn.model_selection._search import BaseSearchCV
 
 import openml
 from ..tasks import get_task