of built-in training loops such as `tf.keras` `compile` and `fit`, we
        expect reduction value to be `SUM` or `NONE`. Using `AUTO` in that case
        will raise an error.
    *   `NONE`: Weighted losses with one dimension reduced (axis=-1, or axis
        specified by loss function). When this reduction type used with built-in

  Depending on the value returned from QueueingHintFn, the scheduling queue changes how it queues each Pod:
  - if more than one QueueingHintFn returns QueueImmediately, it queues Pod to activeQ.
  - If no QueueingHintFn returns QueueImmediately and more than one plugin returns QueueAfterBackoff, it queues Pod to backoffQ if Pod is backing off, or to activeQ if Pod's backoff has already finished.

ains","a","b","bup","compareDocumentPosition","rcssescape","fcssescape","ch","asCodePoint","charCodeAt","escapeSelector","sel","preferredDoc","pushNative","Expr","outermostContext","sortInput","hasDuplicate","documentIsHTML","rbuggyQSA","dirruns","done","classCache","createCache","tokenCache","compilerCache","nonnativeSelectorCache","sortOrder","booleans","identifier","attributes","pseudos","rwhitespace","rcomma","rleadingCombinator","rdescend","rpseudo","ridentifier","matchExpr","ID","CLASS","T...

* Fixed race condition issue when detaching azure disk, preventing `Multi-Attach error`s when scheduling one pod from one node to another. ([#60183](https://github.com/kubernetes/kubernetes/pull/60183), [@andyzhangx](https://github.com/andyzhangx))

### StatefulSet

* Fix zone placement heuristics so that multiple mounts in a StatefulSet pod are created in the same zone ([#40910](https://github.com/kubernetes/kubernetes/pull/40910), [@justinsb](https://github.com/justinsb))

ains","a","b","bup","compareDocumentPosition","rcssescape","fcssescape","ch","asCodePoint","charCodeAt","escapeSelector","sel","preferredDoc","pushNative","Expr","outermostContext","sortInput","hasDuplicate","documentIsHTML","rbuggyQSA","dirruns","done","classCache","createCache","tokenCache","compilerCache","nonnativeSelectorCache","sortOrder","booleans","identifier","attributes","pseudos","rwhitespace","rcomma","rleadingCombinator","rdescend","rpseudo","ridentifier","matchExpr","ID","CLASS","T...

<= 1.32 proceeded to schedule pods onto nodes with no devices by not allocating any devices for those pods. Kubernetes 1.33 changes that to only picking nodes which have at least one device. Users who want the "proceed with scheduling also without devices" semantic can use the upcoming prioritized list feature with one sub-request for "all" devices and a second alternative with "count: 0". ([#129560](https://github.com/kubernetes/kubernetes/pull/129560), [@bart0sh](https://github.com/bart0sh)) [SIG...

the time specified by `--legacy-service-account-token-clean-up-period` (defaulting to one year), **and** are referenced from the `.secrets` list of a ServiceAccount object, **and**  are not referenced from pods. This label causes the authentication layer to reject use of the credentials. After being labeled as invalid, if the time specified by `--legacy-service-account-token-clean-up-period` (defaulting to one year) passes without the credential being used, the secret is automatically deleted. Secrets...

- `kubectl run --env` no longer supports CSV parsing. To provide multiple environment
  variables, use the `--env` flag multiple times instead. Example: `--env ONE=1 --env TWO=2` instead of `--env ONE=1,TWO=2`.

- Removed deprecated command `kubectl stop`.

- Kubectl can now use http caching for the OpenAPI schema. The cache

    * --alpha-endpoint-reconciler-type=(master-count, lease, none) (default
    * "master-count"). The original reconciler is 'master-count'. The 'lease'
    * reconciler uses the storageapi and a TTL to keep alive an endpoint within the
    * `kube-apiserver-endpoint` storage namespace. The 'none' reconciler is a noop
    * reconciler that does not do anything. This is useful for self-hosted