Skip to main content

How to Create a Notebook

Inside JupyterLab, creating a notebook means starting a kernel session connected to a compute resource (CPU or GPU). This is separate from provisioning the AI Notebook instance on the portal.

Create a CPU notebook​

  1. In the Launcher, click the + button.
  2. Choose CPU Notebook.

Once ready, the status bar at the bottom shows CPU (free) | idle.

Create a GPU notebook​

  1. Ensure your Workspace has enough credits to cover at least one hour of GPU cost.
  2. In the Launcher, choose a notebook with your preferred GPU (e.g. 1× GPU H100 SXM5).
  3. Click Confirm on the confirmation dialog.
  4. Wait 1–2 minutes for the notebook to be ready.

Once ready, the status bar shows GPU flavor name | idle.

Idle behavior and auto-shutdown​

AI Notebook automatically disconnects idle sessions to prevent unexpected charges.

TriggerIdle thresholdWhat happens
GPU Kernel idle30 minutes of no code executionKernel is automatically disconnected.
Lab Environment idle1 hour of no interaction (no typing, clicking, or tab switching)Lab is automatically deleted to free resources.
note

If the lab is deleted due to idleness, your saved files remain in persistent storage. Only unsaved work (in-memory variables, unsaved editor buffers) is lost. Save frequently.

Billing​

Billing for the GPU instance starts the moment the notebook on the portal reaches Running — not when you start a kernel session inside JupyterLab. It stops only when you explicitly stop the notebook from the portal, not when you close the browser or log out.