Update app.py

app.py

@@ -7,7 +7,9 @@ import gc
 @st.cache_resource
 def load_model():
     model_name = "peterxyz/detect-llama-34b"
-    tokenizer = LlamaTokenizer.from_pretrained(model_name)
+    
+    # Load the specific tokenizer type used by the model checkpoint
+    tokenizer = AutoTokenizer.from_pretrained(model_name, use_fast=False)
     
     # Check if CUDA is available
     device = "cuda" if torch.cuda.is_available() else "cpu"