Skip to main content

dstack

After installing dstack and starting Transformer Lab, follow these steps to add it as a compute provider.

Add dstack provider

Add dstack in Team Settings​

  1. Open Team Settings by clicking your username in the sidebar.
  2. Go to Compute Providers.
  3. Click Add Compute Provider.
  4. In the modal:
    • Set Type to dstack (beta).
    • Give the provider a name (e.g. dstack-provider).
    • Fill in the Server URL, API Token, and Project Name fields.
  5. Click Add Compute Provider.

You can also add the provider via the CLI with lab provider add.

Run health check​

After the provider is listed in Team Settings:

  1. Find your dstack provider in Compute Providers.
  2. Click the "Check provider status" icon (heartbeat) next to your dstack provider in the status column.
  3. Confirm the provider reports as active.

dstack-specific behavior​

  • Transformer Lab maps one run/cluster to one dstack run.
  • You can target a dstack fleet via fleet_name in the yaml resources field.

Troubleshooting​

  • Setting up backends: To use a specific backend on dstack, see the dstack backends documentation.
  • Provider check fails: Verify server_url reachability from the API host and confirm token validity.
  • Run creation fails: Check resource availability on your dstack dashboard at http://<dstack-server-url>/offers.
  • First launch failure: When setting up a new dstack server, create a fleet first — otherwise the first run will fail with no hardware offers available. Go to your dstack dashboard, navigate to the Fleets tab, and create a fleet with the desired hardware configuration.