Update app.py

app.py

@@ -362,8 +362,8 @@ model_ids = [
     "sentence-transformers/distiluse-base-multilingual-cased-v2",
     "Alibaba-NLP/gte-multilingual-base",
     "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2",
+    "jinaai/jina-reranker-v2-base-multilingual",
     "BAAI/bge-reranker-v2-m3",
-    "jinaai/jina-reranker-v2-base-multilingual"
 ]
 # model_id = "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2"
 # model_id = "Alibaba-NLP/gte-multilingual-base"
@@ -373,20 +373,28 @@ model_ids = [
 # model_id = "sentence-transformers/distiluse-base-multilingual-cased-v2"
 
 model_id = model_ids[-1]
+model = None
+codes_emb = None
 
-if model_id in model_ids[-2:]:
-    model = CrossEncoder(
-        # "jinaai/jina-reranker-v2-base-multilingual",
-        "BAAI/bge-reranker-v2-m3",
-        automodel_args={"torch_dtype": "auto"},
-        trust_remote_code=True,
-    )
-else:
-    model = SentenceTransformer(model_id, trust_remote_code=True)
+def load_model(model_id):
+    global model
+    global codes_emb
+    if model_id in model_ids[-2:]:
+        model = CrossEncoder(
+            # "jinaai/jina-reranker-v2-base-multilingual",
+            # "BAAI/bge-reranker-v2-m3",
+            model_id,
+            automodel_args={"torch_dtype": "auto"},
+            trust_remote_code=True,
+        )
+    else:
+        model = SentenceTransformer(model_id, trust_remote_code=True)
+        # codes_emb = model.encode([x[6:] for x in codes])
+        codes_emb = model.encode([x["examples"] for x in examples])#.mean(axis=1)
+        # codes_emb = np.mean([model.encode(x["examples"]) for x in examples], axis=1)
+        
+load_model(model_id)
 
-# codes_emb = model.encode([x[6:] for x in codes])
-codes_emb = model.encode([x["examples"] for x in examples])#.mean(axis=1)
-# codes_emb = np.mean([model.encode(x["examples"]) for x in examples], axis=1)
 # for x in examples:
 #   codes_emb.append(model.encode(x["examples"]))
 # codes_emb = np.mean(codes_emb, axis=1)
@@ -690,10 +698,11 @@ def reload(chosen_model_id):
     global codes_emb
 
     if chosen_model_id != model_id:
-        model = SentenceTransformer(chosen_model_id, trust_remote_code=True)
-        model_id = chosen_model_id
+        load_model(model_id)
+        # model = SentenceTransformer(chosen_model_id, trust_remote_code=True)
+        # model_id = chosen_model_id
         # codes_emb = model.encode([x[6:] for x in codes])
-        codes_emb = model.encode([x["examples"] for x in examples])#.mean(axis=1)
+        # codes_emb = model.encode([x["examples"] for x in examples])#.mean(axis=1)
         # codes_emb = np.mean([model.encode(x["examples"]) for x in examples], axis=1)
         return f"Model {chosen_model_id} has been succesfully loaded!"
     return f"Model {chosen_model_id} is ready!"