* Switch k8s on GCE to use GCI by default ([#33353](https://github.com/kubernetes/kubernetes/pull/33353), [@vishh](https://github.com/vishh))
* Move HighWaterMark to the top of the struct in order to fix arm, second time ([#33376](https://github.com/kubernetes/kubernetes/pull/33376), [@luxas](https://github.com/luxas))

for an example.

*  TensorFlow IO support is now available for Apple Silicon packages.

*  Refactor CpuExecutable to propagate LLVM errors.

## Keras

Keras is a framework built on top of the TensorFlow. See more details on the Keras [website](https://keras.io/).

### Major Features and Improvements

* `tf.keras`

of Heapster. Instead, you may use the [metrics server](https://github.com/kubernetes-incubator/metrics-server). It's a simplified heapster that is able to gather and serve current metrics values. It provides the Metrics API that is used by `kubectl top`, and horizontal pod autoscaler. Note that it doesn't include some features of Heapster, such as short term metrics for graphs in kube-dashboard and dedicated push sinks, which proved hard to maintain and scale. Clusters using Heapster for transferring...

* Implement NetworkPolicy v1beta1 API object / client support. ([#25638](https://github.com/kubernetes/kubernetes/pull/25638), [@caseydavenport](https://github.com/caseydavenport))
* Only expose top N images in `NodeStatus` ([#25328](https://github.com/kubernetes/kubernetes/pull/25328), [@resouer](https://github.com/resouer))

* Fix fluentd-gcp configuration to facilitate JSON parsing ([#48139](https://github.com/kubernetes/kubernetes/pull/48139), [@crassirostris](https://github.com/crassirostris))
* Bump runc to v1.0.0-rc2-49-gd223e2a - fixes `failed to initialise top level QOS containers` kubelet error. ([#48117](https://github.com/kubernetes/kubernetes/pull/48117), [@sjenning](https://github.com/sjenning))

- Fixed a bug where the kubelet ephemerally failed with `failed to initialize top level QOS containers: root container [kubepods] doesn't exist`, due to the cpuset cgroup being deleted on cgroup v2 with systemd cgroup manager.

- Added a `--show-swap` option to `kubectl top` subcommands ([#129458](https://github.com/kubernetes/kubernetes/pull/129458), [@iholder101](https://github.com/iholder101)) [SIG CLI]

	for _, fivs := range versionsMap {
		// Removal of existing versions and adding a delete marker in the same
		// request is supported. At the same time, we cannot allow adding
		// two delete markers on top of any object. To avoid this situation,
		// we will sort deletions to execute existing deletion first,
		// then add only one delete marker if requested
		sort.SliceStable(fivs.Versions, func(i, j int) bool {

#include "tensorflow/core/platform/thread_annotations.h"
#include "tensorflow/core/platform/types.h"
#include "tensorflow/core/public/release_version.h"
#include "tensorflow/core/public/session.h"

// The implementation below is at the top level instead of the
// brain namespace because we are defining 'extern "C"' functions.
using tensorflow::AttrValueMap;
using tensorflow::DataType;
using tensorflow::ExtendSessionGraphHelper;
using tensorflow::FullTypeDef;

	}

	nodeCollector = newMinioCollectorNode(nodeGroups)
	clusterCollector = newMinioClusterCollector(allMetricsGroups)
	bucketCollector = newMinioBucketCollector(bucketMetricsGroups)
}

// MetricNamespace is top level grouping of metrics to create the metric name.
type MetricNamespace string

// MetricSubsystem is the sub grouping for metrics within a namespace.
type MetricSubsystem string

const (