VESSL Storage is the default, fully managed storage system provided by the VESSL platform. It is automatically created for each organization and cannot be deleted, ensuring continuous availability for all projects.

​
Creating a volume in VESSL Storage

Volume is the basic storage unit in VESSL. There are two ways to create a volume, each optimized for different machine learning workflow scenarios.

In YAML, a volume in VESSL Storage is specified as volume://vessl-storage/{volume-name}.

​
1. User-generated volume

Users can create volumes to upload datasets or models that will be used in workloads.

1

Create a new volume

Navigate to Storage > VESSL Storage and click the + Create a new volume button. Specify the volume name and click the Create button.

2

Upload files to the volume

Click the Add files button in the volume and select the files in your local machine to upload.

3

Import the volume to a run

When creating a run in the web console, create a new directory (e.g., /dataset), and import the volume of VESSL Storage as shown below.

Alternatively, you can manually specify the volume in YAML editor as follows:

import:
    /dataset: volume://vessl-storage/my-volume

​
2. Workload-generated volume

Volumes can also be created by a workload when exporting logs, checkpoints, and metrics to VESSL Storage. The volume name is automatically generated in this case.

1

Export a workload volume to storage

To export your workload’s data to VESSL Storage, create a new directory (e.g., /output), and export to VESSL Storage as follows.

Alternatively, you can manually specify the export in the YAML editor as follows.

 export:
     /output: volume://vessl-storage/
2

Locate the exported volume

Navigate to Run > Files to find the file tree of run execution.

Follow the export volume link to access the volume in the VESSL Storage.

Whether your volume is user-generated or workload-generated, tagging allows for quick search and retrieval of data.

OverviewExternal Storage