Datasets
Create datasets to use as training and validation sets for fine-tuning.
Create a dataset in the **AI Infrastructure Rental** workspace (chosen in **Switch Platform**). The **Create Fine-Tuning Task** form has a required **Training Set** field and a required **Validation Set** field, and both pickers show no entries on a tenant that has no dataset.
The [**Datasets**](https://console.zettabyte.com/library-datasets/list) list has a **Show Only My** toggle and the columns **Name/ID**, **Size**, **Status**, **Upload Progress**, **Latest Version**, **Updated Time** and **User**.
Before you begin
With **Shared Storage** selected, **Datasets Address** lists files that are already in shared storage, so place the data there before creating the dataset. See [Storage and PVCs](https://docs.zettabyte.com/docs/zcloud/models-data-storage/storage) and [Bring your own model or dataset](https://docs.zettabyte.com/docs/zcloud/models-data-storage/bring-your-own-model-dataset).
Create a dataset
- Open [**Datasets**](https://console.zettabyte.com/library-datasets/list) and select **Create Dataset**.
- In **Dataset Name** (required), enter a name. The field's hint reads "Lowercase start, alphanumeric and hyphens."
- Optionally, enter a **Description**.
- Under **Storage Type** (required), keep **Shared Storage**, which is selected when the form opens, or select **S3 Storage**.
- With **Shared Storage**, select an entry in **Datasets Address** (required), a list of files under the shared-data path, each with a path and a date, and a **Refresh** control beside it. With **S3 Storage**, enter **Access Key**, **Secret Key**, **Region** and **S3 Address** in place of **Datasets Address**.
- Optionally, enter a **Version Description**.
- Select **Create Dataset**. The new dataset appears as a row in the **Datasets** list, with **Latest Version** V1.
The **Version** field shows `v1`.
Next steps
[Fine-tune a model](https://docs.zettabyte.com/docs/zcloud/tasks/fine-tune-a-model): use a dataset as the training set and the validation set.