Presets
Training presets are read-only named configurations that combine a runtime, a sizing, and optional defaults. They let you start a training job in a few clicks instead of filling out the full form.
Training presets are managed by your admin via a ConfigMap. You can't create or edit them in the UI, you can only use them to pre-fill the job creation form.
The presets list
The list view shows three sections:
Named presets
Named presets combine a runtime reference with a sizing key. Click Use this preset to open the training job creation form with the runtime and sizing pre-selected.
| Column | What it shows |
|---|---|
| Key | The preset's identifier (e.g. pytorch-single-gpu). |
| Runtime | The training runtime this preset uses. |
| Sizing | The compute sizing this preset uses. |
| Description | What this preset is for. |
Sizing presets
Sizing presets define the compute shape for a training job, nodes, CPU, memory, and GPU per node. They're referenced by named presets but can also be used directly in the custom job form.
| Column | What it shows |
|---|---|
| Key | The sizing identifier (e.g. small-1x-gpu). |
| Display name | Human-readable label. |
| Nodes | Number of nodes. |
| CPU / Memory | Per-node resources. |
| GPU | GPU type and count per node, if any. |
TTL and defaults
Default time-to-live values for completed and failed jobs, plus default CPU and memory per node.
Next steps
- Submitting a job: how to use a preset when creating a job
- Runtimes: the training environments presets reference