Skip to main content
Datasets provide input data for development environments and training jobs. You can select the data you need from the datasets provided by Siverse and mount it for reading when you create a workload.

​
Basic concepts

​
Select a dataset

Before selecting a dataset, confirm that it is available to the workload in the current region and cluster.

​
Usage flow

  1. Confirm that the target dataset is available to the workload in the current region and cluster.
  2. When creating a development environment or training job, select a dataset, specify a subdirectory as needed, and configure the container mount path.
  3. Read data in your code through the mount path. Write training outputs, checkpoints, Notebooks, or temporary files to a writable Volume.
  4. If the same dataset will be read repeatedly and the project already uses dataset caching, you can configure a cache for it.

​
Read and write boundaries

Datasets are typically used as input data. Mounting a dataset does not automatically modify scripts or convert data formats. The data path in your code must match the container mount path. If your program needs to write output files, cache files, training artifacts, or model files, use a writable Volume. Do not write outputs to the dataset mount path or mount a dataset over a critical directory provided by the image.