Update app.py

app.py

@@ -1,6 +1,6 @@
 import os
 import gradio as gr
-from transformers import AutoTokenizer, AutoModelForCausalLM
+from transformers import LlamaTokenizer, AutoModelForCausalLM
 import torch
 import json
 from datetime import datetime
@@ -10,7 +10,7 @@ HUGGINGFACE_TOKEN = os.getenv("HUGGINGFACE_TOKEN")
 
 # Load Llama 3.2 (QLoRA) Model on CPU
 MODEL_NAME = "meta-llama/Llama-3.2-1B-Instruct-QLORA_INT4_EO8"
-tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, token=HUGGINGFACE_TOKEN)
+tokenizer = LlamaTokenizer.from_pretrained(MODEL_NAME, token=HUGGINGFACE_TOKEN)
 model = AutoModelForCausalLM.from_pretrained(
     MODEL_NAME,
     token=HUGGINGFACE_TOKEN,
@@ -19,7 +19,7 @@ model = AutoModelForCausalLM.from_pretrained(
 
 # Load Llama Guard for content moderation on CPU
 LLAMA_GUARD_NAME = "meta-llama/Llama-Guard-3-1B-INT4"
-guard_tokenizer = AutoTokenizer.from_pretrained(LLAMA_GUARD_NAME, token=HUGGINGFACE_TOKEN)
+guard_tokenizer = LlamaTokenizer.from_pretrained(LLAMA_GUARD_NAME, token=HUGGINGFACE_TOKEN)
 guard_model = AutoModelForCausalLM.from_pretrained(
     LLAMA_GUARD_NAME,
     token=HUGGINGFACE_TOKEN,