In aiWARE, the end-to-end processing of an AI workload is called a job. Jobs are composed of one or more tasks. A task specifies an engine, a payload, the task's priority, due dates, and input and output folders. When creating a job, an order of operations is defined, from data ingestion to engine processing to output. Output from one task can be subsequently processed by another engine as part of another task. The paths along which the data flows can be represented by a Directed Acyclic Graph (DAG), which is a tree of tasks.
In aiWARE, these are called Job DAGs. Jobs operate on a Temporal Data Object (TDO), which is a generic data wrapper that stores metadata about media files.
There are two types of jobs in aiWARE:
- Ad hoc jobs that run processing on a single file
- Scheduled jobs that are executed on a schedule
You can use aiWARE's Data Center to easily create jobs, and then use Processing Center to monitor jobs and their tasks, view task output, and cancel pending jobs.
This documentation covers:
