How to Manage GPU Kernels
Go to Running Kernels / Notebooks in the navigation bar (second icon from the top).
Running GPU Kernel table
Displays all currently running GPU kernels in your lab.
| Field | Description |
|---|---|
| Name | Unique identifier of the GPU kernel. |
| Created At | Date and time when the GPU kernel started. |
| Action | Shut Down — terminates the selected kernel. Shut Down All — terminates all running kernels. Terminated kernels move to the Kernel History section. |
Kernel History table
Displays all terminated GPU kernels. Use this to verify kernel usage duration and cross-check with billing data on the Portal.
| Field | Description |
|---|---|
| Name | Unique identifier of the GPU kernel. |
| Created At | Date and time when the GPU kernel started. |
| Duration | Total runtime of the GPU kernel. |
Billing reconciliation
You may notice a slight difference — typically 1–10 seconds — between the GPU runtime shown in AI Factory Billing and the Duration column in Kernel History. This is expected due to normal synchronization delays between internal services.
ghi chú
Stopping a notebook from the portal terminates all running kernels. Save your work in JupyterLab before stopping. See Stop, Start, and Delete for portal-level controls.