Debug prints

aggregate_server_json.py

@@ -204,13 +204,15 @@ class AggregateNER:
             mean_score = 0
         #sum_deviation = 0
         #for node in dist:
-        #    sum_deviation += (mean_score - node["confidence"])*(mean_score - node["confidence"])
+         #   sum_deviation += (mean_score - node["confidence"])*(mean_score - node["confidence"])
         #variance = sum_deviation/len(dist)
         #std_dev = math.sqrt(variance)
-        #threshold =  mean_score + std_dev*self.threshold #default is 1 standard deviation from mean
-        threshold = mean_score
+        #print("mean",mean,"std_dev",std_dev)
+        #threshold =  mean_score - std_dev*self.threshold #default is 1 standard deviation from mean
+        threshold = mean_score 
         pick_count = 1
         for node in dist:
+            print("conf:",node["confidence"],"threshold:",threshold)
             if (node["confidence"] >= threshold):
                 ret_arr.append({"e":node["e"],"conf":node["confidence"]})
                 pick_count += 1
@@ -292,8 +294,10 @@ class AggregateNER:
                         #here cs and ci are same. So use two consecutive cs predictions if meaningful
                         if (len(results[server_index]["orig_cs_prediction_details"][pivot_index]['cs_distribution']) >= 2):
                             ret_arr = self.get_predictions_above_threshold(results[server_index]["orig_cs_prediction_details"][pivot_index])
+                            print("orig cs:",results[server_index]["orig_cs_prediction_details"][pivot_index]['cs_distribution'])
                             orig_cs_second_entity = results[server_index]["orig_cs_prediction_details"][pivot_index]['cs_distribution'][1]
                             m2_cs = orig_cs_second_entity["e"].split('[')[0]
+                            print("m2_cs",m2_cs,"ret_arr",ret_arr)
                             is_cs_included = True if (m2_cs in servers_arr[server_index]["precedence"]) else False
                             is_cs_included = True #Disabling cs included check. If prediction above threshold is cross prediction, then letting it through
                             assert (m2_cs != m1)