This topic describes how to automatically import Machine Learning Platform for AI (PAI) models to Object Storage Service (OSS) for a scheduling task.

Procedure

  1. Log on to the PAI console and choose Model Training > Visualized Modeling (Machine Learning Studio). For more information, see Use DataWorks tasks to schedule experiments in Machine Learning Studio.
    When you create a project, we recommend that you select the pay-as-you-go billing method and enable GPU-accelerated computing. PAI-TensorFlow tasks can run only by using GPU resources.
  2. In the left-side navigation pane of the page that appears, click Models. On the page that appears, drag the model that you want to manage to the canvas, right-click the model, and then choose Model Option > Show Model to obtain the model name.
    PAI-Studio
  3. In the dialog box that appears, find the model name and remove the prefix pailr_coefficient from the name.
    Delete a prefix
  4. Determine the Alibaba Cloud resource name (ARN) that is the unique credential used by PAI to access OSS.
    OSS ARN
  5. Run the command to save the model.
    To periodically import models to OSS, you must first create an SQL node in DataWorks and run the following command:
    pai -name modeltransfer2oss
        -DmodelName=xlab_m_logisticregres_1006580_v0(Model name)
        -DossPath="oss://images-beijing.oss-cn-beijing-internal.aliyuncs.com/aohai/"(OSS path)
        -Darn="acs:ram::1664081855183111:role/aliyunodpspaidefaultrole"(ARN)
        -Dformat=pmml
        -Doverwrite=true/false;