- K8s.io/apimachinery - scheme.Convert() now uses only explicitly registered conversions - default reflection based conversion is no longer available. `+k8s:conversion-gen` tags can be used with the `k8s.io/code-generator` component to generate conversions. ([#90018](https://github.com/kubernetes/kubernetes/pull/90018), [@wojtek-t](https://github.com/wojtek-t))...

    `tf.train.experimental.enable_mixed_precision_graph_rewrite()`.
*   Add environment variable `TF_CUDNN_DETERMINISTIC`. Setting to "true" or "1"
    forces the selection of deterministic cuDNN convolution and max-pooling
    algorithms. When this is enabled, the algorithm selection procedure itself
    is also deterministic.

## Breaking Changes

*   Many backwards incompatible API changes have been made to clean up the APIs

to match the specification. It's unfortunate // that we need to do a shift instead of a mask, but see the comment on // rightShift. // // These are the most dangerous lines in the package and maybe in the // library: a single bit of bias in the selection of nonces would likely // lead to key recovery, but no tests would fail. Look but DO NOT TOUCH. if excess := len(b)*8 - c.N.BitLen(); excess > 0 { // Just to be safe, assert that this only happens for the one curve that // doesn't have a round number...