Refactor code

rag.py

@@ -28,6 +28,8 @@ chat_llm = HuggingFaceEndpoint(
     do_sample=False,
 )
 
+global_tokenizer = LlamaTokenizer.from_pretrained("meta-llama/Meta-Llama-3-8B-Instruct")
+
 
 def local_retriever(query: str):
     topChunks = 3
@@ -274,9 +276,7 @@ def global_retriever(query: str, level: int, response_type: str):
         i += 1
 
     ###Debug####
-    global_tokenizer = LlamaTokenizer.from_pretrained(
-        "meta-llama/Meta-Llama-3-8B-Instruct"
-    )
+
     tokens = global_tokenizer.encode(intermediate_results)
     print(f"Number of input tokens: {len(tokens)}")
     ###Debug###