运行命令时:
az ml service create realtime -f score.py --model-file model.hdf5 -s schema.json -n modelapp -r python --collect-model-data false -c aml_config\conda_dependencies.yml
图像和服务似乎都已创建。但是过了一会儿,这个错误出现了:
{
"Azure-cli-ml Version": "0.1.0a27.post3",
"Error": {
"Error Message": "No response from health endpoint after multiple deploy attempts. Setting status to failed."
},
"Response Code": 500,
"Response Content": {
"CreatedTime": "2018-02-16T14:40:56.358161Z",
"EndTime": "2018-02-16T14:51:23.79374Z",
"Error": {
"Code": "DeploymentFailed",
"Message": "No response from health endpoint after multiple deploy attempts. Setting status to failed.",
"StatusCode": 500
},
"OperationType": "Service",
"State": "Failed"
}
}
当我运行az ml service logs realtime
命令时,我看到很多消息说
2018-02-16T14:59:44.964990Z, INFO, 00000000-0000-0000-0000-000000000000, , 127.0.0.1 - - [16/Feb/2018:14:59:44 +0000] "GET / HTTP/1.0" 200 7 "-" "Go-http-client/1.1"
我正在使用 Azure ML Workbench 版本 0.1.1712.18263。大约一个月前,当我第一次将端点部署到此模型管理帐户时,此过程运行良好。在 Azure 中,我看到该服务的状态为:失败且没有 URL,但已填充主键和辅助键。
我尝试重新安装该软件以防万一它是旧版本,但它没有帮助。我还能做些什么来使端点处于活动状态?