Update app.py

app.py

@@ -10,20 +10,20 @@ from langchain.memory import ConversationBufferMemory
 from langchain_community.chat_models import ChatLlamaCpp
 
 
-callbacks = [StreamingStdOutCallbackHandler()]
-print("creating ll started")
-llm = ChatLlamaCpp(
-    model_path="finbro-v0.1.0-llama-3-8B-instruct-1m.gguf",
-    n_batch=8,
-    temperature=0.85,
-    max_tokens=256,
-    top_p=0.95,
-    top_k = 10,
-    callback_manager=callbacks,
-    n_ctx=2048,
-    verbose=True,  # Verbose is required to pass to the callback manager
-)
-print("creating llm ended")
+# callbacks = [StreamingStdOutCallbackHandler()]
+# print("creating ll started")
+# llm = ChatLlamaCpp(
+#     model_path="finbro-v0.1.0-llama-3-8B-instruct-1m.gguf",
+#     n_batch=8,
+#     temperature=0.85,
+#     max_tokens=256,
+#     top_p=0.95,
+#     top_k = 10,
+#     callback_manager=callbacks,
+#     n_ctx=2048,
+#     verbose=True,  # Verbose is required to pass to the callback manager
+# )
+# print("creating llm ended")