Update app.py

app.py

@@ -16,7 +16,7 @@ tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL)
 
 # Force CPU-safe model loading
 base_model = AutoModelForCausalLM.from_pretrained(BASE_MODEL, torch_dtype=torch.float32)
-model = PeftModel.from_pretrained(base_model, ADAPTER_PATH, device_map=None).to_empty(device='cpu')
+model = PeftModel.from_pretrained(base_model, ADAPTER_PATH, device_map=None).to("cpu")
 
 
 # Dictionaries to decode user inputs