CLIGPU Deployments
hudl gpu deploy --offer <id>
Provision a new GPU cluster against a chosen offer.
Provision a new bare-metal GPU cluster. The --offer flag is required and must reference a valid offer ID from hudl gpu offers. The deployment process is asynchronous -- the command returns immediately and the cluster transitions through provisioning states until it is ready.
Usage
hudl gpu deploy --offer <id> [flags]Flags
| Flag | Description |
|---|---|
--offer <id> | (required) The GPU offer ID to deploy against. |
--region <code> | Override the active region for this call only. |
--workspace <slug> | Override the active workspace for this call only. |
-o, --output | Output format: table (default), json, yaml, wide, name. |
--idempotency-key <key> | Attach an idempotency key so retries are safe. |
Examples
Deploy a cluster using an offer ID obtained from hudl gpu offers:
hudl gpu deploy --offer off_8xH100_SXM5_eu2Deploy to a specific region and output the result as JSON:
hudl gpu deploy --offer off_8xH100_SXM5_us1 --region us-east-1 -o jsonSafe retry with an idempotency key (prevents double-provisioning):
hudl gpu deploy --offer off_4xA100_80G_eu2 --idempotency-key "deploy-train-job-42"API reference
This command calls POST /gpu/deployments under the hood. See the full request and response schema in the API reference.