Help Center> ModelArts> User Guide (Senior AI Engineers)> Migrating the Data of MLS to ModelArts

Migrating the Data of MLS to ModelArts

HUAWEI CLOUD plans to terminate Machine Learning Service (MLS). You are advised to use related functions of MLS in ML Studio of ModelArts. To ensure the normal running of service data, you are advised to perform the following operations to migrate data.

Migrating Data

The MLS pro edition supports data to be stored in HDFS, and the MLS standard edition supports data to be stored in datasets and OBS. ML Studio reads data from and writes data to OBS. Therefore, you need to manually migrate MLS data to OBS.

For details about how to create an OBS bucket and folder, see Creating a Bucket and Creating a Folder.

For details about how to upload files to OBS, see Uploading a File.

To facilitate service migration to ModelArts, you are advised to create the same OBS buckets and folders as those of the data stored in MLS.

Create buckets and folders in OBS, and upload the data and models of MLS to the new OBS buckets.

Migrating Workflows

The workflow capabilities of ML Studio and MLS are the same. The supported data source is changed. As a result, the operators for reading and writing data are changed. You need to orchestrate the operators in the workflow again. For details about operator changes, see Table 1 and Table 2. Read Model From OBS and Save Model to OBS only support the models in inner format.

Table 1 Changes between operators in the MLS pro edition and ML Studio

Category

Operator of MLS Pro Edition

Operator of ML Studio

Input

Read Model

Read Model From OBS

Read PMML

Not supported

Read Hive

Not supported

Read HDFS

Read Data From OBS

Read Text

Read Text From OBS

Output

Save HDFS

Save Data to OBS

Save PMML

Not supported

Save Model

Save Model to OBS

Save Hive

Not supported

Table 2 Changes between operators in the MLS standard edition and ML Studio

Category

Operator of MLS Standard Edition

Operator of ML Studio

Input

Read Data from CloudTable

Not supported

Read Data From Dataset

Read Data From OBS

Read Model

Read Model From OBS

Read Text From Dataset

Read Text From OBS

Output

Save Data to Dataset

Save Data to OBS

Save Model

Save Model to OBS

Migrating Notebooks

ModelArts Notebook provides you with better coding experience. You can download the notebook script of MLS to the local PC and upload the script to ModelArts for further use. In this case, you must migrate notebook data and code.

Migrating Notebook Data

The MLS pro edition saves notebook data in HDFS, and the MLS standard edition saves notebook data in OBS. Upload the notebook data to OBS. For details, see Migrating Data.

Migrating Notebook Code

The procedure is as follows:

  1. Log in to the ModelArts management console. In the left navigation pane, choose DevEnviron > Notebooks to switch to the Notebooks page.
  2. Click Create. The Create Notebook page is displayed. Set Work Environment to Multi-Engine 1.0 (Python3, Recommended) and Storage to OBS. For details about how to set other parameters, see Creating a Notebook Instance.
  3. Download the notebook script in .ipynb format from MLS.
    Figure 1 Downloading notebook scripts
  4. Upload the .ipynb file to a notebook development environment.
    Figure 2 Uploading the script to a notebook
  5. Click to open the .ipynb file, and set the kernel to Conda-python2. Click Set Kernel to complete the setting. Then, you can use the .ipynb file.
    Figure 3 Setting the kernel

For details about how to operate files in OBS in a notebook development environment, see MoXing-Common Component.