ARVICA / DOCUMENTATION

Docs / Workloads

Images and GPU environments

Make the software stack part of your deployment requirements.

Select a template

Use the template catalogue as a starting point and review what the requested environment contains. A template selection records intent; it does not prove an image has been installed on a host.

Compatibility checklist

Specify the operating system, CPU architecture, driver requirements, CUDA runtime and framework version. GH200 uses an Arm-based host architecture, so confirm that your images and dependencies support the target host.

Validate after activation

Verify device visibility, available GPU memory and framework GPU detection before a long run. Run a small workload with the intended precision and batch size, then save the working environment versions with your project.

Explore templates