Update chain_setup.py

chain_setup.py

@@ -15,12 +15,13 @@ def load_llm():
     )
 
     llm = LlamaCpp(
-        model_path=model_file,
-        flash_attn=False,
-        n_ctx=2048,
-        n_batch=512,
-        chat_format="chatml"
-    )
+    model_path=model_file,
+    flash_attn=False,
+    n_ctx=2048,  # or 4096
+    n_batch=512, # or even 256
+    chat_format='chatml'
+)
+
     return llm
 
 def build_conversational_chain(vectorstore):