Quick Deploy

User

Quick Deploy sits at the top of the Jobs page: a row of prepared templates - PyTorch, TensorFlow, a chat UI - already configured with an image and the ports they need.

If you are new, start here. You give it a runtime and nothing else.

Running one

1. Find the template. Filter templates… narrows the list; the cards page with the arrows on the right.
2. Read More shows the full description and the ports it declares.
3. Deploy asks for a runtime and shows the estimated cost.
4. Run it.

The minimum runtime is five minutes; below that the run action stays disabled.

Editing one first

Edit on a template loads it into the normal submission form instead of running it. You get the template’s image and ports as a starting point and can then mount folders, change the queue, or set environment variables.

This is the useful middle path: the template knows which image and ports go together, and you supply the parts that are specific to you.

Templates do not mount anything for you

A deployed template has no storage mounted, so its output is discarded when the runtime ends. If you intend to keep results, use Edit, mount a bucket, and run it from the form.

Where templates come from

Templates are configured for the tenant, not per user. If one you want is missing, that is a request to your tenant administrator.

Dismissing the Quick Deploy panel with the × only hides it for the session.