Update chain_setup.py

chain_setup.py

@@ -17,11 +17,12 @@ def load_llm():
     llm = LlamaCpp(
     model_path=model_file,
     flash_attn=False,
-    n_ctx=2048,  # or 4096
-    n_batch=512, # or even 256
-    chat_format='chatml'
+    n_ctx=2048,
+    n_batch=512,
+    chat_format="chatml",
+    grammar=""  # set grammar to an empty string to bypass template parsing issues
 )
-
+    
     return llm
 
 def build_conversational_chain(vectorstore):