Update app.py

app.py

@@ -22,7 +22,7 @@ TOKEN = os.environ.get("TOKEN", None)
 
 OWNER="Blanca"
 DATA_DATASET = f"{OWNER}/CQs-Gen_test_embeddings"
-INTERNAL_DATA_DATASET = f"{OWNER}/CQs-Gen_test_embeddings
+INTERNAL_DATA_DATASET = f"{OWNER}/CQs-Gen_test_embeddings"
 SUBMISSION_DATASET = f"{OWNER}/submissions_internal"
 SUBMISSION_DATASET_PUBLIC = f"{OWNER}/submissions_public"
 #CONTACT_DATASET = f"{OWNER}/contact_info" # TODO: I should reactivate this
@@ -217,14 +217,12 @@ def add_new_eval(
         with open(file_path, 'r') as f:
             data = json.load(f)
             scores = []
-            for id_to_eval, line in data.items():
+            for id_to_eval, line in data.items(): # data to evaluate
                 intervention_score = 0
-                for indx, intervention_id in enumerate(gold_dataset['intervention_id']):
-                    if indx == 3:
-                        break
+                for indx, intervention_id in enumerate(gold_dataset['intervention_id']): # references
                     if id_to_eval == intervention_id:
                         references = gold_dataset['cqs']
-                        reference_set = [row['cq'] for row in references[indx]]
+                        reference_embeddings = [row['embedding'] for row in references[indx]]
                         # TODO: here upload the embedding that I have saved, so they can be used in similarity evaluation
                         
                         #print(reference_set, flush=True)
@@ -235,8 +233,8 @@ def add_new_eval(
                             
                             if METRIC == 'similarity':
                                 sentence_embedding = similarity_model.encode(cq_text)
-                                reference_embedding = similarity_model.encode(reference_set) # TODO: here have the embeddings directly, do no calculate each time
-                                sims = similarity_model.similarity(sentence_embedding, reference_embedding).tolist()[0]
+                                #reference_embedding = similarity_model.encode(reference_set) # TODO: here have the embeddings directly, do no calculate each time
+                                sims = similarity_model.similarity(sentence_embedding, reference_embeddings).tolist()[0]
     
                                 winner = np.argmax(sims)
                                 # make sure the similarity of the winning reference sentence is at least 0.65