Update app.py

app.py

@@ -29,6 +29,7 @@ SUBMISSION_DATASET_PUBLIC = f"{OWNER}/submissions_public"
 #CONTACT_DATASET = f"{OWNER}/contact_info"
 RESULTS_DATASET = f"{OWNER}/results_public"
 LEADERBOARD_PATH = f"HiTZ/Critical_Questions_Leaderboard"
+METRIC = 'similarity'
 api = HfApi()
 
 YEAR_VERSION = "2025"
@@ -162,28 +163,29 @@ def add_new_eval(
                     if id_to_eval == intervention_id:
                         references = gold_dataset['cqs']
                         reference_set = [row['cq'] for row in references[indx]]
-                        print(reference_set, flush=True)
+                        #print(reference_set, flush=True)
                         for cq in line['cqs']:
                             # TODO: compare to each reference and get a value
                             cq_text = cq['cq']
-                            print(cq_text, flush=True)
-
-                            #if args.metric == 'similarity':
-                            sentence_embedding = similarity_model.encode(cq_text)
-                            reference_embedding = similarity_model.encode(reference_set)
-                            sims = similarity_model.similarity(sentence_embedding, reference_embedding).tolist()[0]
-                            print(sims, flush=True)
-
-                            winner = np.argmax(sims)
-                            # make sure the similarity of the winning reference sentence is at least 0.65
-                            if sims[winner] > 0.65:
-                                label = references[indx][winner]['label']
-                                if label == 'Useful':
-                                    score += 1/3
+                            #print(cq_text, flush=True)
+
+                            
+                            if METRIC == 'similarity':
+                                sentence_embedding = similarity_model.encode(cq_text)
+                                reference_embedding = similarity_model.encode(reference_set)
+                                sims = similarity_model.similarity(sentence_embedding, reference_embedding).tolist()[0]
+                                #print(sims, flush=True)
+    
+                                winner = np.argmax(sims)
+                                # make sure the similarity of the winning reference sentence is at least 0.65
+                                if sims[winner] > 0.65:
+                                    label = references[indx][winner]['label']
+                                    if label == 'Useful':
+                                        score += 1/3
                             #else: 
                             #    label = 'not_able_to_evaluate'
-                
-                return format_error(score)
+                print(indx, score, flush=True)
+                #return format_error(score)