* Add external link [Machine learning model serving in Python using FastAPI and Streamlit](https://davidefiocco.github.io/2020/06/27/streamlit-fastapi-ml-serving.html) to docs. PR [#1669](https://github.com/tiangolo/fastapi/pull/1669) by [@davidefiocco](https://github.com/davidefiocco).

*   `Estimator.export_savedmodel()` now includes all valid serving signatures
    that can be constructed from the Serving Input Receiver and all available
    ExportOutputs. For instance, a classifier may provide regression- and
    prediction-flavored outputs, in addition to the classification-flavored one.
    Building signatures from these allows TF Serving to honor requests using the

- Kubeadm: add the deprecated flag --port=0 to kube-controller-manager and kube-scheduler manifests to disable insecure serving. Without this flag the components by default serve (e.g. /metrics) insecurely on the default node interface (controlled by --address). Users that wish to override this behavior and enable insecure serving can pass a custom --port=X via kubeadm's "extraArgs" mechanic for these components. ([#92720](https://github.com/kubernetes/kubernetes/pull/92720),...

above. panic("crypto/ecdh: internal error: public key is the identity element") } // A "Pairwise Consistency Test" makes no sense if we just generated the // public key from an ephemeral private key. Moreover, there is no way to // check it aside from redoing the exact same computation again. SP 800-56A // Rev. 3, Section 5.6.2.1.4 acknowledges that, and doesn't require it. // However, ISO 19790:2012, Section 7.10.3.3 has a blanket requirement for a // PCT for all generated keys (AS10.35) and FIPS 140-3...