Refactoring #105.

common/experiment/extract/opinions.py

@@ -82,7 +82,7 @@ def fill_opinion_collection(collection, linked_data_iter, labels_helper, to_opin
 
         agg_label = labels_helper.aggregate_labels(
             labels_list=list(linked.iter_labels()),
-            label_creation_mode=label_calc_mode)
+            label_calc_mode=label_calc_mode)
 
         agg_opinion = to_opinion_func(linked.First, agg_label)
 

common/model/labeling/base.py

@@ -16,7 +16,7 @@ def label_to_uint(self, label):
     def get_classes_count(self):
         return len(self._label_scaler.ordered_suppoted_labels())
 
-    def aggregate_labels(self, labels_list, label_creation_mode):
+    def aggregate_labels(self, labels_list, label_calc_mode):
         raise NotImplementedError()
 
     @staticmethod

common/model/labeling/modes.py

@@ -1,17 +1,8 @@
-# TODO: Use enum instead.
-class LabelCalculationMode:
+from enum import Enum
 
-    FIRST_APPEARED = u'take_first_appeared'
-    AVERAGE = u'average'
 
-    @staticmethod
-    def supported(value):
-        for s in LabelCalculationMode.__iter_supported():
-            if s == value:
-                return True
-        return False
+class LabelCalculationMode(Enum):
 
-    @staticmethod
-    def __iter_supported():
-        for var_name in dir(LabelCalculationMode):
-            yield getattr(LabelCalculationMode, var_name)
+    FIRST_APPEARED = u'take_first_appeared'
+
+    AVERAGE = u'average'

common/model/labeling/single.py

@@ -6,17 +6,16 @@
 
 class SingleLabelsHelper(LabelsHelper):
 
-    def aggregate_labels(self, labels_list, label_creation_mode):
+    def aggregate_labels(self, labels_list, label_calc_mode):
         assert(isinstance(labels_list, list))
-        assert(isinstance(label_creation_mode, unicode))
-        assert(LabelCalculationMode.supported(label_creation_mode))
+        assert(label_calc_mode, LabelCalculationMode)
 
         label = None
 
-        if label_creation_mode == LabelCalculationMode.FIRST_APPEARED:
+        if label_calc_mode == LabelCalculationMode.FIRST_APPEARED:
             label = labels_list[0]
 
-        if label_creation_mode == LabelCalculationMode.AVERAGE:
+        if label_calc_mode == LabelCalculationMode.AVERAGE:
             int_labels = [self._label_scaler.label_to_int(label)
                           for label in labels_list]
             label = self._label_scaler.int_to_label(np.sign(sum(int_labels)))

contrib/networks/core/callback/utils_model_eval.py

@@ -26,7 +26,8 @@
 
 
 def evaluate_model(experiment, data_type, epoch_index, model,
-                   labels_formatter, save_hidden_params, log_dir):
+                   labels_formatter, save_hidden_params,
+                   label_calc_mode, log_dir):
     """ Performs Model Evaluation on a particular state (i.e. epoch),
         for a particular data type.
     """
@@ -72,6 +73,7 @@ def evaluate_model(experiment, data_type, epoch_index, model,
         data_type=data_type,
         epoch_index=epoch_index,
         result_filepath=result_filepath,
+        label_calc_mode=label_calc_mode,
         labels_formatter=labels_formatter)
 
     # Evaluate.
@@ -91,14 +93,15 @@ def evaluate_model(experiment, data_type, epoch_index, model,
 
 def __convert_output_to_opinion_collections(exp_io, opin_ops, doc_ops, labels_scaler, opin_fmt,
                                             result_filepath, data_type, epoch_index,
-                                            labels_formatter):
+                                            label_calc_mode, labels_formatter):
     assert(isinstance(opin_ops, OpinionOperations))
     assert(isinstance(doc_ops, DocumentOperations))
     assert(isinstance(labels_scaler, BaseLabelScaler))
     assert(isinstance(exp_io, NetworkIOUtils))
     assert(isinstance(opin_fmt, OpinionCollectionsFormatter))
     assert(isinstance(data_type, DataType))
     assert(isinstance(epoch_index, int))
+    assert(isinstance(label_calc_mode, LabelCalculationMode))
     assert(isinstance(labels_formatter, StringLabelsFormatter))
 
     opinions_source = exp_io.get_input_opinions_filepath(data_type=data_type)
@@ -112,8 +115,7 @@ def __convert_output_to_opinion_collections(exp_io, opin_ops, doc_ops, labels_sc
         labels_scaler=labels_scaler,
         create_opinion_collection_func=opin_ops.create_opinion_collection,
         keep_doc_id_func=lambda doc_id: doc_id in cmp_doc_ids_set,
-        # TODO. bring this onto parameters level.
-        label_calculation_mode=LabelCalculationMode.AVERAGE,
+        label_calculation_mode=label_calc_mode,
         output=MulticlassOutput(labels_scaler=labels_scaler,
                                 has_output_header=True))