Dataset Management

Creating Datasets and Uploading Data

See Data Upload.

Binding Datasets to Containers

When creating a container, you can select the created dataset in bind data to bind it to a specified directory in the container.

When binding data, you need to select the directory to be mapped in the container. As shown in the figure below, binding the cifar_10 data to /input0 means that the content of the dataset can be accessed through the /input0 directory in the container.

For example, if the cifar10 dataset contains a file meta.csv, and it is bound to the /input0 directory in the container, then the file can be read through the path /input0/meta.csv.

The binding directory of a specific container can also be viewed in the container details, as shown below:

Changing Types

On the settings page of "Datasets", you can change the type of the dataset:

After changing it to "Model", it will be displayed in the data warehouse of "Model" type.

Deleting Data Versions and Entire Datasets

As mentioned above, a dataset can contain multiple versions. If a specific version is no longer needed, it can be deleted using the "Delete Current Version" button on the right side of the "Versions" list. After the version is deleted, the storage resources it occupied will be released.

Of course, if the entire dataset is no longer needed, it can also be permanently deleted. After deletion, all versions under the dataset will be deleted and will no longer occupy the user's storage quota.