TF_AddInput(desc, {zero, 0});
  TF_AddInput(desc, {input, 0});
  TF_SetAttrInt(desc, "num_split", 3);
  TF_SetAttrType(desc, "T", TF_INT32);
  // Set device to CPU since there is no version of split for int32 on GPU
  // TODO(iga): Convert all these helpers and tests to use floats because
  // they are usually available on GPUs. After doing this, remove TF_SetDevice
  // call in c_api_function_test.cc
  TF_SetDevice(desc, "/cpu:0");

* Do not query the metadata server to find out if running on GCE.  Retry metadata server query for gcr if running on gce. ([#28871](https://github.com/kubernetes/kubernetes/pull/28871), [@vishh](https://github.com/vishh))
* Fix GPU resource validation ([#28743](https://github.com/kubernetes/kubernetes/pull/28743), [@therc](https://github.com/therc))

    * Fix stop hook failure on kubernetes-worker charm

    * Fix handling of juju kubernetes-worker.restart-needed state

    * Fix nagios checks in charms

  * Enable GPU mode if GPU hardware detected ([#43467](https://github.com/kubernetes/kubernetes/pull/43467), [@tvansteenburgh](https://github.com/tvansteenburgh))

*   GCE nodes with NVIDIA GPUs attached now expose nvidia.com/gpu as a resource instead of alpha.kubernetes.io/nvidia-gpu. ([#54826](https://github.com/kubernetes/kubernetes/pull/54826),[ @mindprince](https://github.com/mindprince))

* Support for the `alpha.kubernetes.io/nvidia-gpu` resource, which was deprecated in 1.10, has been removed. Please use the resource exposed by DevicePlugins instead (`nvidia.com/gpu`). ([#61498](https://github.com/kubernetes/kubernetes/pull/61498), [@mindprince](https://github.com/mindprince))

    for (auto input : inputs) {
      TF_AddInput(desc, input);
    }
    // Set device to CPU because some ops inside the function might not be
    // available on GPU.
    TF_SetDevice(desc, "/cpu:0");
    *op = TF_FinishOperation(desc, s_);
    ASSERT_EQ(TF_OK, TF_GetCode(s_)) << TF_Message(s_);
    ASSERT_NE(*op, nullptr);
  }

  FunctionDef fdef() {

- Adds the ability to disable Accelerator/GPU metrics collected by Kubelet ([#91930](https://github.com/kubernetes/kubernetes/pull/91930), [@RenaudWasTaken](https://github.com/RenaudWasTaken)) [SIG Node]

- sigs.k8s.io/kustomize/api: v0.8.5
- sigs.k8s.io/kustomize/cmd/config: v0.9.7
- sigs.k8s.io/kustomize/kustomize/v4: v4.0.5
- sigs.k8s.io/kustomize/kyaml: v0.10.15

### Changed
- dmitri.shuralyov.com/gpu/mtl: 666a987 → 28db891
- github.com/Azure/go-autorest/autorest: [v0.11.1 → v0.11.12](https://github.com/Azure/go-autorest/autorest/compare/v0.11.1...v0.11.12)

  - Base pricing of custom instances on their instance family type
  - Add pricing information for missing machine types
  - Add pricing information for different GPU types
  - Ignore the new `topology.gke.io/zone` label when comparing groups
  - Add missing stable labels to template nodes
  
  ### HuaweiCloud
  - Add auto scaling group support
  - Implement node group by AS

* List all nodes and occupy cidr map before starting allocations ([#29062](https://github.com/kubernetes/kubernetes/pull/29062), [@bprashanth](https://github.com/bprashanth))
* Fix GPU resource validation ([#28743](https://github.com/kubernetes/kubernetes/pull/28743), [@therc](https://github.com/therc))