Skip to main content
Preset datasets are public datasets maintained by Siverse. You can select and mount one directly when creating a development environment or training job. Before use, confirm that the preset dataset’s region, cluster, and visibility scope meet the requirements of your task.

​
Use preset datasets

  1. Create a development environment or training job.
  2. Select the target preset dataset in the dataset mount configuration.
  3. If you need only part of the dataset, configure a subdirectory.
  4. Specify the dataset mount path inside the container. Use a separate directory to avoid overwriting image-provided directories such as /opt, /workspace, or /usr.
  5. Submit the create or update request.

​
Verify read results

After the development environment or training job starts, go to the mount path in the container and confirm that your code can read the dataset contents. Datasets are typically mounted as read-only inputs. If your code needs to write output files, cache files, or training artifacts, write them to a writable Volume. If you cannot find the dataset contents, first check that the mount path is correct. If you mounted only a subdirectory, check that your code reads from the corresponding path inside the container.

​
Next steps