We have included a new model that copes with sentences in Spanish.

main_service.py

@@ -1,8 +1,7 @@
 import gradio as gr
 from utils_model import ModelFactory
 
-def retrieve_embeddings(input_text_query):
-    model_type = 'all-MiniLM-L6-v2'
+def retrieve_embeddings(input_text_query, model_type):
     model_factory_obj = ModelFactory()
     model = model_factory_obj.create_model(model_type)
 
@@ -11,7 +10,11 @@ def retrieve_embeddings(input_text_query):
     return query_embeddings
 
 def build():
-    app = gr.Interface(fn=retrieve_embeddings, inputs="text", outputs="dataframe")
+    models_list = ['all-MiniLM-L6-v2', 'sentence_similarity_spanish']
+
+    app = gr.Interface(fn=retrieve_embeddings, 
+                       inputs=["text", gr.Dropdown(models_list, label='Model type')], 
+                       outputs="dataframe")
 
     return app
 

utils_model.py

@@ -10,12 +10,12 @@ class ModelFactory():
     def create_model(self, model_type):
         model = None
 
-        if (model_type=='mock'):
-            model = MockModel()
-
         if (model_type=='all-MiniLM-L6-v2'):
             model = MiniLM_L6_v2_Model()
 
+        if (model_type=='sentence_similarity_spanish'):
+            model = SentenceSimilaritySpanishModel()
+
         return model
 
 class BaseModel():
@@ -24,26 +24,18 @@ class BaseModel():
         pass
 
     def retrieve_embeddings(self, input_text):
-        pass    
-
-class MockModel(BaseModel):
-
-    def __init__(self):
-        pass
-
-    def retrieve_embeddings(self, input_text):
-        random_embeddings = np.random.randint(256, size=(370))/256
+        embeddings = self.model.encode(input_text, batch_size=32)
+        embeddings *= 255
+        embeddings = embeddings.astype(np.uint8).tolist()
 
-        return pd.DataFrame(random_embeddings)
+        return embeddings
 
 class MiniLM_L6_v2_Model(BaseModel):
     
     def __init__(self):
         self.model = SentenceTransformer('all-MiniLM-L6-v2')
 
-    def retrieve_embeddings(self, input_text):
-        embeddings = self.model.encode(input_text, batch_size=32)
-        embeddings *= 255
-        embeddings = embeddings.astype(np.uint8).tolist()
+class SentenceSimilaritySpanishModel(BaseModel):
 
-        return embeddings
+    def __init__(self):
+        self.model = SentenceTransformer('hiiamsid/sentence_similarity_spanish_es')