Dataset Versions

Creating a New Version

Each time a dataset is successfully created, a new version will be added incrementally. You can view all uploaded data versions in the "Versions" tab on the page.

If you find that the dataset version remains in "Processing" status after upload, it may be because there was a problem with dataset extraction. In this case, you can try re-uploading the data, and the dataset will increment a new version for the new upload. Failed dataset uploads can be deleted directly.

Updating Data in a Dataset Version

:::caution Note Currently, "Dataset" does not support deleting files in a specific version (deleting a single file). You can only delete unwanted files by uploading a new version. :::

Dataset versions currently support incremental data updates. As shown in the image below, clicking "Upload to Current Directory" allows you to update the current directory of the current dataset, which also supports automatic extraction of various compressed packages.

If a file with the same filename exists in the current directory, the newly uploaded file will overwrite the content of the old file.

Data upload and synchronization require processing time depending on the size of the uploaded data. During processing, the current dataset is marked as "Processing," and datasets in this state cannot be bound to containers.

Switching Dataset Versions

After opening the dataset page, the latest version's data content is displayed by default. Click on a version number in the "Versions" tab to switch the dataset version to the selected version, then click "Overview" to see the dataset's file list updated to the current version's content.

You can also select and switch versions in the version tab in the upper right corner.

Previewing Files in a Dataset Version

The dataset basic information displays the file list of the currently selected version. By default, it shows the root directory's file content.

Directories in the list can be expanded to browse files in subdirectories. Some file formats (such as .txt .csv .py .md .yaml .yml .png .jpeg .jpg .ipynb) support preview.

:::info Directory preview displays up to 200 files (or folders). For more data, please view it after "Binding Dataset" to a container. :::

Downloading Data

When "Previewing Files," you can download individual files by clicking the download icon on the right. You can also download the entire dataset version as a zip archive by clicking "Download Current Directory."

:::caution Note For very large-scale datasets, download time may take several hours or even days, and any network fluctuations may cause the dataset download to fail. Downloading is not recommended. If downloading is necessary, it is recommended to download only essential files. :::