none
Continuous Integration in Machine learning server RRS feed

  • Question

  • Hi,

    I have published an R script as a web service using the mrsdeploy package for R  in machine learning server

    But when i publish the web service using the function publishService as given below

    api <- publishService( serviceName, code = manualTransmission, model = carsModel, inputs = list(hp = "numeric", wt = "numeric"), outputs = list(answer = "numeric"), v = "v1.0.0" )

    What ever i assign to the the argument "code" will be stored  in the api data and will be executed whenever the API is consumed.

    Is there anyway to implement the Continuous Integration, Currently what i am doing is

    Code="source('C:/Analytics/RModuleScripts/Rfile1.r'), that is using the source function in r

    so that each time we call the API the web service will execute that particular file from that particular directory

    Is there any better way to implement Continuous Integration 

    Wednesday, July 4, 2018 7:15 AM

Answers

  • Hi Mathew,

    You can pass a filename path in the code argument, see more details here:

    https://docs.microsoft.com/en-us/machine-learning-server/operationalize/how-to-deploy-web-service-publish-manage-in-r#publish-and-update-web-services

    You can also update the service whenever a new file is available. Another option is to call the APIs directly as part of CI: 

    https://docs.microsoft.com/en-us/machine-learning-server/operationalize/concept-api

    We are working on adding az CLI commands for services for our next release so automating CI will be even easier. 

    Thursday, July 5, 2018 4:56 PM