### Major Features and Improvements

* `tf.keras`
    * `Model.compile` now support `steps_per_execution='auto'` as a parameter, allowing automatic tuning of steps per execution during `Model.fit`,
    `Model.predict`, and `Model.evaluate` for a significant performance boost.

## Thanks to our Contributors

    Arg<TF_ResourceTensor, [{A handle to the multi device iterator to delete.}], [TF_DatasetIteratorFree]>:$multi_device_iterator,
    Arg<Variadic<TF_ResourceTensor>, [{A list of iterator handles (unused). This is added so that automatic control dependencies get added during function tracing that ensure this op runs after all the dependent iterators are deleted.}], [TF_DatasetIteratorRead]>:$iterators,
    Arg<TF_VariantTensor, [{A variant deleter.}]>:$deleter
  );

                            may result in more than one call from the gateway to the destination backend,
                            for example, if automatic retries are supported.


                            Because the Request timeout encompasses the BackendRequest timeout, the value of

```

The support for `seccomp.security.alpha.kubernetes.io/pod` and `container.seccomp.security.alpha.kubernetes.io/...` annotations are now deprecated, and will be removed in Kubernetes v1.22.0. Right now, an automatic version skew handling will convert the new field into the annotations and vice versa. This means there is no action required for converting existing workloads in a cluster.