在本地使用docker部署图像分类cntk的训练模型 [英] To deploy the train model of image classification cntk with docker locally

问题描述

部署的整个过程:

(1)

cd C:\ Users \ Zhong \ AppData \ Local \ Temp \ azureml_runs \ classification-CNTK_1534903355816

cd部署

az登录

要登录，请使用网络浏览器打开页面https://microsoft.com/devicelogin，然后输入代码GWVHZ2PQJ进行身份验证.

[

  {

    "cloudName":"AzureCloud"，

    "id":"da6972b0-62c1-40b2-a9d6-089c9a33ab48"，

    "isDefault":true，

    名称":"\ u514d \ u8d39 \ u8bd5 \ u7528"，

    状态":"Enabled"，

    "tenantId":"7ffdc7a0-965c-4f2b-965b-d93ba7600bb8"，

    用户":{

     名称":"[e-mail]"，

     类型":用户"

    }

  }

]

(2)

docker登录

使用现有凭据进行身份验证...

登录成功

&s

(3)az ml env show

{

  集群名称":"zhong"，

  群集大小":"N/A"，

  创建时间":"2018-08-19T11:04:12.123Z"，

  当前模式":本地"，

  位置":"southeastasia"，

  设置状态":成功"，

  资源组":"resource-group"，

  订阅":"da6972b0-62c1-40b2-a9d6-089c9a33ab48"

}

&s

(4)az ml env local

现在以本地模式运行

&s

(5)az ml服务创建实时-c. ./aml_config/conda_dependencies.yml -f deploymain.py -s deployserviceschema.json -n imgclassapi1 -v -r python -d helpers.py -d helpers_cntk.py -d实用程序_CVbasic_v2.py -d实用程序_general_v2.py -d svm.np -d lutId2Label.pickle-模型文件cntk_refined.model

&s

开始创建服务

开始创建图片

开始创建清单

上传pipRequirements文件

上传condaEnvFile文件

&s

启动模型注册

  cntk_refined.model

尝试将模型注册到https://southeastasia. modelmanagement.azureml.net/api/subscriptions/da6972b0-62c1-40b2-a9d6-089c9a33ab48/resourceGroups/resource-group/accounts/azure-ml-model-management/models

尝试使用此信息注册模型

&s

成功注册的模型

Id:d52876b494604ba393b835939db6fd7c

更多信息:'az ml模型展示-m d52876b494604ba393b835939db6fd7c '

在C:\ Users \ Zhong \创建新驱动程序AppData \ Local \ Temp \ tmpr4zr_c1m.py

驱动程序已上传到http://mlcrpstgebb45341441d.blob. core.windows.net/amlbdpackages/27997b48-588b-4367-8414-7d60d4174cd6.py?se=2018-09-21T03%3A25%3A39Z&sp=r&sr=b&sv=2016-05-31&sig= dzbvbNTaZfPlKw8gfyCNKgGi1v6sclXKZfxlBxjOKM4％3D

  helpers.py

已将依赖项helpers.py添加到资产.

  helpers_cntk.py

已将依赖项helpers_cntk.py添加到资产中.

  utilities_CVbasic_v2.py

已将依赖项Utility_CVbasic_v2.py添加到资产中.

  utilities_general_v2.py

已将依存Utility_general_v2.py添加到资产中.

  svm.np

已将依赖项svm.np添加到资产中.

  lutId2Label.pickle

为资产添加了依赖项lutId2Label.pickle.

  deploymain.py

已将依赖项deploymain.py添加到资产中.

  deployserviceschema.json

已将依赖项deployserviceschema.json添加到资产中.

清单有效载荷

&s

成功创建的清单

Id:9cda1ba6-6d1e-40a8-8a3f-e176fd47071c

图片有效载荷:{'computeResourceId':'/subscriptions /da6972b0-62c1-40b2-a9d6-089c9a33ab48/resourcegroups/resource-group/providers/Microsoft.MachineLearningCompute/operationalizationClusters/zhong'， 'imageType':'Docker'，'name':'imgclassapi1'，'description':''，'manifestId':'9cda1ba6-6d1e-40a8-8a3f-e176fd47071c'}

操作ID:55fd11b2-4bd2-40fc-801c- 3084b69fefe8

 

创建图片..... .......................完成.

图片ID:9e3050e5-4c5a-4790-a6d9- c5849f85271a

更多详细信息:'az ml image show -i 9e3050e5 -4c5a-4790-a6d9-c5849f85271a'

用法信息:'az ml图像用法-i 9e3050e5 -4c5a-4790-a6d9-c5849f85271a'

[本地模式]正在运行Docker容器.

[本地模式]从mlcrpacr8fb197558338.azurecr中提取图像.io/imgclassapi1:18.这可能需要几分钟，具体取决于您的连接速度...

[本地模式]拉动中…… .........................................

容器端口:32770

[本地模式]等待容器初始化. .{

    "Azure-cli-ml版本":"0.1.0b2.post2"，

错误": "错误，容器初始化失败.请运行"az ml服务日志实时-i imgclassapi1"以确定原因."

  ;

  ;

  ;

  ;

  ;

  ;

  ;

  ;

  ;

  ;

  ;

  ;

  ;

  ;

  ;

  ;

  ;

  ;

  ;

  ;

  ;

  ;

  ;

  ;

解决方案

请注意，此论坛是在The whole process of deployment:

(1)

cd C:\Users\Zhong\AppData\Local\Temp\azureml_runs\classification-CNTK_1534903355816

cd deploy

 

az login

To sign in, use a web browser to open the page https://microsoft.com/devicelogin and enter the code GWVHZ2PQJ to authenticate.

[

  {

    "cloudName": "AzureCloud",

    "id": "da6972b0-62c1-40b2-a9d6-089c9a33ab48",

    "isDefault": true,

    "name": "\u514d\u8d39\u8bd5\u7528",

    "state": "Enabled",

    "tenantId": "7ffdc7a0-965c-4f2b-965b-d93ba7600bb8",

    "user": {

      "name": "[e-mail]",

      "type": "user"

    }

  }

]

(2)

docker login

Authenticating with existing credentials...

Login Succeeded

 

(3)az ml env show

{

  "Cluster Name": "zhong",

  "Cluster Size": "N/A",

  "Created On": "2018-08-19T11:04:12.123Z",

  "Current Mode": "local",

  "Location": "southeastasia",

  "Provisioning State": "Succeeded",

  "Resource Group": "resource-group",

  "Subscription": "da6972b0-62c1-40b2-a9d6-089c9a33ab48"

}

 

(4)az ml env local

Now running in local mode

 

(5)az ml service create realtime -c ../aml_config/conda_dependencies.yml -f deploymain.py -s deployserviceschema.json -n imgclassapi1 -v -r python -d helpers.py -d helpers_cntk.py -d utilities_CVbasic_v2.py -d utilities_general_v2.py -d svm.np -d lutId2Label.pickle --model-file cntk_refined.model

 

Starting service create

Starting image create

Starting manifest create

Uploading pipRequirements file

Uploading condaEnvFile file

 

Starting model register

 cntk_refined.model

Attempting to register model to https://southeastasia.modelmanagement.azureml.net/api/subscriptions/da6972b0-62c1-40b2-a9d6-089c9a33ab48/resourceGroups/resource-group/accounts/azure-ml-model-management/models

Attempting to register model with this information

 

Successfully registered model

Id: d52876b494604ba393b835939db6fd7c

More information: 'az ml model show -m d52876b494604ba393b835939db6fd7c'

Creating new driver at C:\Users\Zhong\AppData\Local\Temp\tmpr4zr_c1m.py

Driver uploaded to http://mlcrpstgebb45341441d.blob.core.windows.net/amlbdpackages/27997b48-588b-4367-8414-7d60d4174cd6.py?se=2018-09-21T03%3A25%3A39Z&sp=r&sr=b&sv=2016-05-31&sig=dzbvbNTaZfPlKw8gfyCNKgGi1v6sclXKZfxlBxjOKM4%3D

 helpers.py

Added dependency helpers.py to assets.

 helpers_cntk.py

Added dependency helpers_cntk.py to assets.

 utilities_CVbasic_v2.py

Added dependency utilities_CVbasic_v2.py to assets.

 utilities_general_v2.py

Added dependency utilities_general_v2.py to assets.

 svm.np

Added dependency svm.np to assets.

 lutId2Label.pickle

Added dependency lutId2Label.pickle to assets.

 deploymain.py

Added dependency deploymain.py to assets.

 deployserviceschema.json

Added dependency deployserviceschema.json to assets.

Manifest payload

 

Successfully created manifest

Id: 9cda1ba6-6d1e-40a8-8a3f-e176fd47071c

Image payload: {'computeResourceId': '/subscriptions/da6972b0-62c1-40b2-a9d6-089c9a33ab48/resourcegroups/resource-group/providers/Microsoft.MachineLearningCompute/operationalizationClusters/zhong', 'imageType': 'Docker', 'name': 'imgclassapi1', 'description': '', 'manifestId': '9cda1ba6-6d1e-40a8-8a3f-e176fd47071c'}

Operation Id: 55fd11b2-4bd2-40fc-801c-3084b69fefe8

 

Creating image............................Done.

Image ID: 9e3050e5-4c5a-4790-a6d9-c5849f85271a

More details: 'az ml image show -i 9e3050e5-4c5a-4790-a6d9-c5849f85271a'

Usage information: 'az ml image usage -i 9e3050e5-4c5a-4790-a6d9-c5849f85271a'

[Local mode] Running docker container.

[Local mode] Pulling the image from mlcrpacr8fb197558338.azurecr.io/imgclassapi1:18. This may take a few minutes, depending on your connection speed...

[Local mode] Pulling...............................................

Container port: 32770

[Local mode] Waiting for container to initialize...{

    "Azure-cli-ml Version": "0.1.0b2.post2",

"Error": "Error, container failed to initialize. Please run 'az ml service logs realtime -i imgclassapi1' to determine the cause."

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

解决方案

Note that this forum was recreated on https://social.msdn.microsoft.com/Forums/en-US/5ff623a0-2768-43e1-84e8-81657d60c985/to-deploy-the-train-model-of-image-classification-cntk-with-docker-locally?forum=MachineLearning 

Please disregard / close this forum and refer to the above link going forward. Thanks. 

这篇关于在本地使用docker部署图像分类cntk的训练模型的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！