ernie_tiny部署paddle serving后，curl返回 request that this server could not understand

[Nicole1983]

paddler serving部署文本匹配的ernie_tiny后，curl -H "Content-Type:application/json" -X POST -d '{"Data":[['世界上什么东西最小', '世界上什么东西最小？']] , "fetch":["prediction"]}' http://172.0.0.1:9292/ernie/prediction，返回
400 Bad Request
Bad Request
The browser (or proxy) sent a request that this server could not understand.

请问是传入参数格式不对吗？还是操作步骤不正确呢？谢谢！
完整的操作步骤如下：

1. 获取docker 镜像 registry.baidubce.com/paddlepaddle/serving:0.5.0-cuda10.2-cudnn8-devel

2. 容器内依次安装 paddle-serving-server-gpu==0.5.0.post102，paddlepaddle-gpu==2.0.0 和paddle_serving_client

3. 将ernin_tiny的训练模型转为paddle serving模型，ernie_tiny_model包含如下文件
   total 350716
   drwxrwxr-x 2 media media 4096 Mar 24 09:43 ./
   drwxrwxr-x 4 media media 4096 Mar 24 09:44 ../
   -rw-rw-r-- 1 media media 656906 Mar 24 09:43 model
   -rw-rw-r-- 1 media media 358454636 Mar 24 09:43 params
   -rw-rw-r-- 1 media media 356 Mar 24 09:43 serving_server_conf.prototxt
   -rw-rw-r-- 1 media media 147 Mar 24 09:43 serving_server_conf.stream.prototxt

4. 将步骤#3中生成的ernie_tiny_model和ernie_tiny_client拷贝到docker/home/workspace/Serving/python/examples/ernie_tiny 中

5. 使用uwsgi启动服务
   from paddle_serving_server.web_service import WebService
   uci_service = WebService(name="ernie")
   uci_service.load_model_config("./ernie_tiny_model")
   uci_service.prepare_server(workdir="./workdir", port=int(9500), device="cpu")
   uci_service.run_rpc_service()
   app_instance = uci_service.get_app_instance()

6. 启动服务 uwsgi --http :9292 --module uwsgi_service:app_instance

[TeslaZhao]

您好，根据您的报错判断，错误原因webserver端口没有开启。

请参考Paddle Serving的web server示例可以参考python/examples/下bert和lac示例，webserver启动采用run_web_service()方法，在bert示例服务启动方法(https://github.com/PaddlePaddle/Serving/blob/develop/python/examples/bert/bert_web_service.py)，打开web端口为int(9500)

bert_service.run_web_service()

[Nicole1983]

已经好用了，谢谢