update npu api (#8172)

deploy/python/infer.py

@@ -868,7 +868,7 @@ def load_predictor(model_dir,
     elif device == 'NPU':
         if config.lite_engine_enabled():
             config.enable_lite_engine()
-        config.enable_npu()
+        config.enable_custom_device('npu')
     else:
         config.disable_gpu()
         config.set_cpu_math_library_num_threads(cpu_threads)