We have modified the implementation to return the dataframe with

the features, associating the set of embeddings with the text it
refers to.

main_service.py

@@ -8,12 +8,8 @@ class GradioAppManager():
         self.model = model_factory_obj.create_model(model_type)
 
     def __retrieve_embeddings__(self, input_queries_df):
-
         queries_list = input_queries_df.values
-        queries_embeddings_list = []
-        for current_query_aux in queries_list:
-            current_query_embeddings = self.model.retrieve_embeddings(current_query_aux)
-            queries_embeddings_list+=current_query_embeddings
+        queries_embeddings_list = self.model.retrieve_embeddings_from_texts_list(queries_list)
 
         return queries_embeddings_list
 

utils_model.py

@@ -23,12 +23,25 @@ class BaseModel():
     def __init__(self):
         pass
 
-    def retrieve_embeddings(self, input_text):
+    def retrieve_embeddings_from_single_input_text(self, input_text):
         embeddings = self.model.encode(input_text, batch_size=32)
         embeddings *= 255
-        embeddings = embeddings.astype(np.uint8).tolist()
+        embeddings = embeddings.astype(np.uint8).astype(str).tolist()
 
         return embeddings
+    
+    def retrieve_embeddings_from_texts_list(self, input_texts_list):
+        all_embeddings_list = []
+        for current_input_text_aux in input_texts_list:
+            embeddings = self.retrieve_embeddings_from_single_input_text(current_input_text_aux)
+            nof_features = len(embeddings[0])
+            all_embeddings_list += [current_input_text_aux.tolist() + embeddings[0]]
+
+        queries_embeddings_df = pd.DataFrame(all_embeddings_list)
+        columns_list = ['text'] + [f'feature_{idx}' for idx in range(0, nof_features)]
+        queries_embeddings_df.columns = columns_list
+
+        return queries_embeddings_df
 
 class MiniLM_L6_v2_Model(BaseModel):