Update app.py

app.py

@@ -15,13 +15,14 @@ dtype = torch.bfloat16
 quantization_config = BitsAndBytesConfig(load_in_8bit=True)
 
 def load_model():
-    tokenizer = AutoTokenizer.from_pretrained(model_id, use_auth_token=huggingface_token)
+    tokenizer = AutoTokenizer.from_pretrained(model_id, token=huggingface_token)
     model = AutoModelForCausalLM.from_pretrained(
         model_id, 
         torch_dtype=dtype, 
-        device_map=device, 
+        device_map="auto",
         quantization_config=quantization_config,
-        use_auth_token=huggingface_token
+        token=huggingface_token,
+        low_cpu_mem_usage=True
     )
     return tokenizer, model