Skip to main content

Manage datasets

Publish, edit, and delete datasets on the Qwen Cloud platform.

The Datasets page lists all datasets in your account. Use the search bar and filters to find specific datasets.

Dataset list

Each dataset entry shows:
  • Dataset Name: The name of the dataset.
  • Type: The task type and algorithm (e.g., Text Generation / SFT).
  • Examples: Number of data entries in the dataset.
  • Import Status: Whether the data was imported successfully.
  • Publish Status: Whether the dataset is published and available for fine-tuning.
  • Created: Timestamp when the dataset was created.
  • Actions: Available operations.

Dataset status

Datasets have two statuses:
  • Draft: The dataset has been created but is not yet available for fine-tuning. You can edit the dataset to add or remove files.
  • Published: The dataset is finalized and available for use in fine-tuning jobs. Published datasets cannot be edited.
You must publish a dataset before you can select it in a fine-tuning job.

Available actions

The actions available depend on the dataset status: Draft datasets:
ActionDescription
EditModify the dataset name or upload additional files.
PublishFinalize the dataset and make it available for fine-tuning.
DeletePermanently remove the dataset from your account.
Published datasets:
ActionDescription
DetailsView dataset details including upload history and data preview.
DeletePermanently remove the dataset from your account.
Published datasets that are actively used by fine-tuning jobs cannot be deleted. Remove the dataset from all jobs before deleting it.

Publish a dataset

After creating a dataset (which saves it as a draft), you must publish it before it can be used for fine-tuning:
  1. Go to the Datasets page.
  2. Find the draft dataset and click Publish in the Actions column.
  3. The dataset status changes to Published and it becomes available in the dataset selector when creating a fine-tuning job.

View dataset details

Click Details on a published dataset to view:
  • Dataset name, type, and algorithm.
  • Upload history and file list.
  • Data preview showing sample entries from the uploaded files.

Next steps