测试国产加速卡寒武纪MLU下，跑ernie的序列标注任务的无法调用mlu加速任务bug，希望官方开发能够答复，非常感谢！！

[suntao2015005848]

1）PaddlePaddle版本：docker环境下的 ，PaddlePaddle-mlu==0.0.0 2) PaddleHub版本：2.1.0 2）系统环境：centos7.6 linux 任务启动日志：

/opt/py37env/lib/python3.7/site-packages/paddle/fluid/framework.py:193: UserWarning: We will fallback into legacy dygraph on NPU/XPU/MLU/IPU/ROCM devices. Because we only support new eager dygraph mode on CPU/GPU currently. 
  "We will fallback into legacy dygraph on NPU/XPU/MLU/IPU/ROCM devices. Because we only support new eager dygraph mode on CPU/GPU currently. "
[2023-05-29 16:37:34,910] [    INFO] - Already cached /root/.paddlenlp/models/ernie-1.0/ernie_v1_chn_base.pdparams
W0529 16:37:34.916133  5170 device_context.cc:48] Please NOTE: device: 0, MLU Compute Capability: 3.0, Driver API Version: 4.20.18, Runtime API Version: 6.0.2, Cnnl API Version: 1.13.0, MluOp API Version: 0.2.0
[2023-05-29 16:37:54,520] [    INFO] - Loaded parameters from /home/suntao_test/v2_nlp_es005/nlp_mod_005/best_model/model.pdparams
 * Serving Flask app 'hub_serving'
 * Debug mode: off
INFO:werkzeug:WARNING: This is a development server. Do not use it in a production deployment. Use a production WSGI server instead.
 * Running on all addresses (0.0.0.0)
 * Running on http://127.0.0.1:8867
INFO:werkzeug:Press CTRL+C to quit
INFO:werkzeug:127.0.0.1 - - [29/May/2023 16:38:04] "POST /predict/NLP_ES005 HTTP/1.1" 200 -
INFO:werkzeug:127.0.0.1 - - [29/May/2023 16:38:05] "POST /predict/NLP_ES005 HTTP/1.1" 200 -

报错信息：

{
    "msg": "The device should not be 'gpu', since PaddlePaddle is not compiled with CUDA", 
    "status": "000", 
    "results": ""
}