GKE Autopilot Clusters with GPUs and TPUs

Google Kubernetes Engine managed DRANET supports both GPUs and TPUs in autopilot clusters. The setup requires creating a Virtual Private Cloud, a custom ComputeClass for the accelerator type, a ResourceClaimTemplate for RDMA or non-RDMA networking, and deploying a workload that references both. The deployment triggers a scale-up with managed DRANET networking.