A workflow is a configurable automated process made up of one or more job, task, or action call. You must create a YAML file to define your workflow configuration.

Workflow kinds

There are two kinds of workflows: live and batch.

Live workflows

Live workflows are executed locally on the developer's machine. They contain a set of job definitions that spawn jobs in the cloud.
Here's an example of a typical job: 1. Executing a Jupyter Notebook server in the cloud on a powerful node with a lot of memory and a high-performant GPU 2. Opening a browser with a Jupyter web client connected to this server

Batch workflows

Batch workflows serve for orchestration of a set of remote tasks that depend on each other. Batch workflows are executed by the main job that manages workflow graphs by spawning required jobs, waiting for their results, and starting dependent tasks when all of their requirements are satisfied. When possible, operations inside the batch runner will be re-ran on failing to ensure maximum batch stability.
Last modified 4mo ago