Introduction to Storage¶

Files are organized inside buckets. UbiOps has a file system that allows you to either create storage buckets directly on UbiOps, or connect to your own storage buckets on Google, AWS, Azure or any other provider that offers S3-compatible object storage. You can have multiple buckets per project and there always is a default bucket in your UbiOps project.

Every file on UbiOps can be re-used in multiple requests in the same project, or used as intermediate storage for passing files between deployments in a pipeline.

The file system makes it easy and efficient to use files in your deployments. It also allows you to pass large files in request input and output. You can use the file function to manage data for requests: storing the input and output for requests for example or as a way of storing persisting data between requests. The file system also combines well with the training functionality: you are able to store your training data, model artifacts, configuration files and evaluation graphs in UbiOps. Every file on UbiOps can be used by other users and other deployment in the same project.

You can find more information about buckets, files, or using files in your deployments here:

• Buckets
• Files
• Using files in deployments

We also provide several How-to's to help you get started with using the UbiOps storage function:

• How-to interact with S3 buckets
• How-to use files for persistent storage
• How-to manage files within your deployment
• How-to stream files from a bucket
• How-to incorporating model checkpointing
• How-to load libraries from Git
• How-to run NVIDIA Triton Inference Server