Acasia Cloud

Acasia Cloud is where your organization sees what its GPU hardware is doing and acts on it. It is one platform for every customer, but what it reports and controls depends on how you hold the hardware.

​

What you see, by customer type

Bare metal customers hold root access and run their own stack on the machines, so Acasia reports at the node level unless the control plane is installed. Marketplace rentals always run the control plane, installed automatically at checkout. See The control plane.

​

Where each task lives

​

How Cloud is organized

All three are scoped to the active organization. Confirm the organization shown in the lower-left corner before creating clusters, deploying models, or generating credentials.

​

Your first hour

1. Confirm the active organization.
2. Review the machines allocated to you and their available capacity.
3. Create a cluster sized to your workload.
4. Deploy a model from the library, or connect to the cluster over SSH.
5. Create an API key and call the endpoint from your application.

​

Start here