My Data

Scenarios

Typical scenarios of My Data:

• Using a dataset to perform smart refining to generate high-quality datasets required for downstream tasks
• Using a dataset for LLM pre-training and fine-tuning to enhance foundational capabilities, while leveraging human preference data to optimize response quality
• Using a dataset as a test set to evaluate model performance and establish baseline assessments of model capabilities

Operation Guide

1. Log in to the ModelArts console.
2. In the navigation pane on the left, choose Asset Management > Data > My Data. You can view all datasets and the list of assets created by you. You can also filter dataset assets by dataset name, data modality, dataset type, status, and creator.
3. Click on the right of the search bar. On the page that appears on the right, set the search bar. Table 1 describes the configurable items.

   Table 1 My data list configuration

   Category	Parameter	Description
   Basic Settings	Text Wrapping	Enabled: Automatically wraps text within cells. Data asset entries will expand vertically to display all information. Disabled: Text will not wrap; data entries will remain on a single line and information may be truncated.
   	Data Columns	Freeze none: If the data list exceeds the screen width, all columns remain scrollable. Freeze first: The first column is frozen while other columns remain scrollable. Freeze first two: The first and second columns are frozen while other columns remain scrollable.
   	Operation Column	When enabled, the Operation column is permanently pinned as the final column and its width cannot be adjusted.
   Custom Columns	Display options	Select the column names you wish to display. Dataset Name and Operation are displayed by default; other columns can be toggled on or off. Column headers can be dragged to reorder them.

4. Select a dataset and perform the following operations in the Operation column:
   • Bring Online: Offline datasets can be brought online. Click Bring Online. In the dialog box that is displayed, confirm the operation. The dataset is brought online. A dataset that has been brought online can be used as data for subsequent development.
   • Take Offline: Online datasets can be taken offline. Click Take Offline. In the dialog box that is displayed, confirm the operation. The dataset is taken offline. A dataset that has been taken offline cannot be used as data for subsequent development.
   • Delete: Datasets can be deleted. Deleted datasets are not permanently removed. If you accidentally delete a dataset, you can restore it for future use. You can permanently delete a deleted dataset. Once permanently deleted, the dataset cannot be restored.
   • Restore: You can restore a deleted dataset.

Data Asset Details Management

The data asset details page displays detailed information about the current dataset. In the dataset workspace, click a dataset name to go to the dataset details page. This page consists of the basic information, data preview, and operation records tab pages. The following describes the functions and operations on the three tab pages. You can delete a dataset in the upper right corner of the page. After you click Delete, the dataset will be permanently deleted. Exercise caution when performing this operation.

• Basic Information Displays the following asset information:
  • Data content: Information such as asset name, modality, type, sample count, size, description, creator, storage location, etc.
  • Data Source: Specifies the task that generated the data asset; you can navigate to the task via the Source Task ID.
  • Extended Info: Attributes and copyright information of the data asset; this information supports manual modification.
    Figure 1 Basic information about data assets
    

• Data Preview: Allows you to display three to five typical samples of structured data (text and tables), view the samples on multiple pages, and view the original data structure. Unstructured data (images/audio) can be previewed in thumbnail mode. Datasets can be downloaded.
• Operation Record: You can view operation records on this page. Operation records include all operations performed on the current dataset.