After a labeling job is complete, export the results as a manifest file to an Object Storage Service (OSS) bucket. You can also create a dataset from the results to use directly for model training in PAI.
Prerequisites
Before you begin, ensure that you have:
A completed labeling job. On the Task Center tab, confirm that the job status is Complete. For more information, see Process labeling jobs.
The workspace administrator or labeling administrator role. If you don't have the required permissions, contact your workspace administrator to assign the labeling administrator role to your account. For more information, see Manage members of a workspace.
Export the results of a labeling job
This procedure exports all labeling results for the selected job.
Go to the iTAG page.
Log on to the PAI console.
In the left-side navigation pane, click Workspaces. On the Workspaces page, click the name of the workspace that you want to manage.
In the left-side navigation pane, choose Data Preparation > iTAG.
On the Task Center tab of the iTAG page, find the labeling job whose results you want to export and click Export Labeling Result in the Actions column.
In the Export Labeling Result dialog box, configure the parameters and click OK.
Parameter Description Export to Dataset Specifies whether to create a dataset from the labeling results. Select Yes if you plan to use the results for model training in PAI — the dataset is stored in the OSS bucket you specify and is available directly on the Dataset management page. Output OSS Directory The OSS bucket path where the labeling results are exported. If Export to Dataset is set to Yes, the dataset is also created in this path. 
After the export completes, check the exported result file. The exported file has a
.manifestsuffix. For details about the file format, see Overview.
Track export progress and access results
Check export progress: In the upper-right corner of the iTAG page, click the export progress icon
. A panel opens showing the status of your export jobs. When the status shows Obtained, click Redirect to open the OSS bucket and view the exported manifest file.
View the created dataset: If you set Export to Dataset to Yes, go to the Dataset management page to view and use the created dataset.
What's next
Use the exported dataset to train models:
Designer overview — build pipelines with a visual drag-and-drop interface
Data Science Workshop (DSW) overview — run interactive notebooks for model development
Before you begin — set up your PAI environment before training