`tensorflow` pip package has a new, optional installation method for Linux that installs necessary Nvidia CUDA libraries through pip. As long as the Nvidia driver is already installed on the system, you may now run `pip install tensorflow[and-cuda]` to install TensorFlow's Nvidia CUDA library dependencies in the Python environment. Aside from the Nvidia driver, no other pre-existing Nvidia CUDA packages are necessary.

*   Enable JIT-compiled i64-indexed kernels on GPU for large tensors with...

	maps.Copy(oi2Map, oi2.UserTags)
	if (oi2.UserTagCount > 0 && !reflect.DeepEqual(oi2Map, t.ToMap())) || (oi2.UserTagCount != len(t.ToMap())) {
		return replicateMetadata
	}

	// Compare only necessary headers
	compareKeys := []string{
		"Expires",
		"Cache-Control",
		"Content-Language",
		"Content-Disposition",
		"X-Amz-Object-Lock-Mode",
		"X-Amz-Object-Lock-Retain-Until-Date",

- Avoid unnecessary GCE API calls when adding IP alises or reflecting them in Node object in GCE cloud provider. ([#90242](https://github.com/kubernetes/kubernetes/pull/90242), [@wojtek-t](https://github.com/wojtek-t)) [SIG Apps, Cloud Provider and Network]

  * After your master is upgraded to 1.6, you will need to update your PodSpecs to set the `tolerations` field of the PodSpec and remove the `scheduler.alpha.kubernetes.io/tolerations` annotation on the Pod. (It's not strictly necessary to remove the annotation, as it will have no effect once you upgrade to 1.6.) Between the time the master is upgraded to 1.6 and when you do this, tolerations attached to Pods that are created will have no effect. Pods that are already running will...

    // maximumSize/Weight is specified in which case ensure that each segment gets at least 10
    // entries. The special casing for size-based eviction is only necessary because that eviction
    // happens per segment instead of globally, so too many segments compared to the maximum size
    // will result in random eviction behavior.
    int segmentShift = 0;
    int segmentCount = 1;

          <type>String</type>
          <description>
            Whether to load Maven extensions (such as packaging and type handlers) from
            this plugin. For performance reasons, this should only be enabled when necessary. Note: While the type
            of this field is {@code String} for technical reasons, the semantic type is actually
            {@code Boolean}. Default value is {@code false}.
          </description>

</li>

<li>
After substitution, each type argument must <a href="#Satisfying_a_type_constraint">satisfy</a>
the <a href="#Type_parameter_declarations">constraint</a> (instantiated, if necessary)
of the corresponding type parameter. Otherwise instantiation fails.
</li>
</ol>

<p>
Instantiating a type results in a new non-generic <a href="#Types">named type</a>;

- **Cluster Lifecycle**
  - [alpha] Ensure critical cluster infrastructure pods (Heapster, DNS, etc.) can schedule by evicting regular pods when necessary to make the critical pods schedule. ([docs](http://kubernetes.io/docs/admin/rescheduler/#guaranteed-scheduling-of-critical-add-on-pods)) ([kubernetes/features#62](https://github.com/kubernetes/enhancements/issues/62))

not present in the json. ([#37991](https://github.com/kubernetes/kubernetes/pull/37991), [@pwittrock](https://github.com/pwittrock))

* The strategicmerge patchMergeKey for VolumeMounts was changed from "name" to "mountPath".  This was necessary because the name field refers to the name of the Volume, and is not a unique key for the VolumeMount.  Multiple VolumeMounts will have the same Volume name if mounting the same volume more than once.  The "mountPath" is verified to be unique and...

- The `SupportNodePidsLimit` feature is now beta, and enabled by default.  It is no longer necessary to set the feature gate `SupportNodePidsLimit=true`. ([#76221](https://github.com/kubernetes/kubernetes/pull/76221), [@RobertKrawitz](https://github.com/RobertKrawitz))