Update app.py

app.py

@@ -13,7 +13,7 @@ def init_retriever():
     Initialize and return the retriever function
     """
     callback_manager = CallbackManager([StreamingStdOutCallbackHandler()])
-    llm = LlamaCpp(model_path="./models/llama-2-13b-chat.Q4_K_S.gguf", 
+    llm = LlamaCpp(model_path="./models/openhermes-2.5-neural-chat-7b-v3-1-7b.Q5_K_M.gguf", 
                    n_ctx=4000, 
                    max_tokens=4000,
                    f16_kv=True,