* elements without resizing. Note that this is not what {@link HashSet#HashSet(int)} does, but it
   * is what most users want and expect it to do.
   *
   * <p>This behavior can't be broadly guaranteed, but has been tested with OpenJDK 1.7 and 1.8.
   *
   * @param expectedSize the number of elements you expect to add to the returned set
   * @return a new, empty hash set with enough capacity to hold {@code expectedSize} elements

* Conformance tests are added for the DaemonSet kinds in the apps/v1 group version. Deprecated versions of DaemonSet will not be tested for conformance, and conformance is only applicable to release 1.10 and later. ([#60456](https://github.com/kubernetes/kubernetes/pull/60456), [@kow3ns](https://github.com/kow3ns))

    *   They are off by default. Enable them by setting the environment variable
        `TF_ENABLE_ONEDNN_OPTS=1`.
    *   We do not recommend using them in GPU systems, as they have not been
        sufficiently tested with GPUs yet.
*   TensorFlow pip packages are now built with CUDA11.2 and cuDNN 8.1.0

## Breaking Changes

*   The `TF_CPP_MIN_VLOG_LEVEL` environment variable has been renamed to