When you run an AutoAI experiment in a project, the type, size, and power of the hardware configuration available depend on the type of experiment you build.
Default hardware configurations
The type of hardware configuration available for your AutoAI experiment depends on the type of experiment you are building. A standard AutoAI experiment, with a single data source, has a single, default hardware configuration. An AutoAI experiment with joined data has options for increasing computational power.
Capacity units per hour for AutoAI experiments
| Capacity type | Capacity units per hour |
|---|---|
| 8 vCPU and 32 GB RAM | 20 |
The runtimes for AutoAI stop automatically when processing is complete.
Compute usage in projects
AutoAI consumes compute resources as CUH from the watsonx.ai Runtime service.
You can monitor the total monthly amount of CUH consumption for the watsonx.ai Runtime service on the Resource usage page on the Manage tab of your project.
What consumes CUH?
Capacity unit hours (CUH) are consumed for running assets, not for working in tools. That is, there is no consumption charge for defining experiments in AutoAI, but there is a charge for running the experiment to train the model candidate pipelines. Similarly, there is no charge for defining a deployment or deployment job for a model created with AutoAI, but there is a charge for running a deployment job or scoring the deployed model.
Learn more
- AutoAI
- watsonx.ai Runtime service
- Compute resource options for assets and deployments in spaces
- Monitoring account resource usage
Parent topic: Choosing compute resources for tools