Deep learning VM images in VMware Private AI Foundation with NVIDIA are distributed in a shared content library published by VMware. As a cloud administrator, you use a content library to pull specific VM images in your VI workload domain during VM deployment.

Prerequisites

As a cloud administrator, verify that VMware Private AI Foundation with NVIDIA is deployed and configured. See Deploying VMware Private AI Foundation with NVIDIA.

Procedure

As a cloud administrator, create a content library with deep learning VM images. If you plan to depoy deep learning VMs by using a YAML file and the kubectl command, you add the content library to the namespace you created for AI workloads.

  1. Log in to the vCenter Server instance for the VI workload domain at https://<vcenter_server_fqdn>/ui.
  2. Select Menu > Content Libraries and click Create.
  3. Create a content library for the deep learning VM images.
    • For a connected environment, create a subscribed content library that is connected to https://packages.vmware.com/dl-vm/lib.json. Authentication is not required.
    • For a disconnected environment, download the deep learning VM images from https://packages.vmware.com/dl-vm/ and import them in to a local content library.

      For each image, download the relevant .ovf, .vmdk, .mf, and .cert files.

    See Create and Edit a Content Library and How to Populate Libraries with Content.

  4. If you plan to enable deployments of deep learning VMs on a Supervisor by directly calling kubectl, add the content library to the vSphere namespace for AI workloads.

    VMware Aria Automation creates a namespace every time a deep learning VM is provisioned, automatically adding the content library to it.

    1. Select Menu > Workload Management.
    2. Navigate to the namespace for AI workloads.
    3. On the VM Service card, click Manage content libraries.
    4. Select the content library with the deep learning VM images and click OK.