*   New `tf.data.Dataset.window()` API allows users to create finite windows
        of input dataset; when combined with the `tf.data.Dataset.reduce()` API,
        this allows users to implement customized batching.
    *   All C++ code moves to the `tensorflow::data` namespace.
    *   Add support for `num_parallel_calls` to `tf.data.Dataset.interleave`.
*   `tf.contrib`:

- Fixed a bug in client-go where new clients with customized `Dial`, `Proxy`, `GetCert` config may get stale HTTP transports. ([#95427](https://github.com/kubernetes/kubernetes/pull/95427), [@roycaihw](https://github.com/roycaihw)) [SIG API Machinery]

-motion-media-query {\n    @media (prefers-reduced-motion: reduce) {\n      transition: none;\n    }\n  }\n}\n","// Form control focus state\n//\n// Generate a customized focus state and for any input with the specified color,\n// which defaults to the `$input-focus-border-color` variable.\n//\n// We highly encourage you to not customize the default value, but instead use\n// this to tweak colors on an as-needed basis. This aesthetic change is based on\n// WebKit's default styles, but applicable...

-motion-media-query {\n    @media (prefers-reduced-motion: reduce) {\n      transition: none;\n    }\n  }\n}\n","// Form control focus state\n//\n// Generate a customized focus state and for any input with the specified color,\n// which defaults to the `$input-focus-border-color` variable.\n//\n// We highly encourage you to not customize the default value, but instead use\n// this to tweak colors on an as-needed basis. This aesthetic change is based on\n// WebKit's default styles, but applicable...