*   Adds `beta` hyperparameter to
    [FTRL](https://www.tensorflow.org/api_docs/python/tf/keras/optimizers/Ftrl)
    optimizer classes (Keras and others) to match
    [FTRL paper](https://research.google.com/pubs/archive/41159.pdf).
*   `Optimizer.__init__` now accepts a `gradient_aggregator` to allow for
    customization of how gradients are aggregated across devices, as well as

- When evicting, Pods in Pending state are removed without checking PDBs. ([#83906](https://github.com/kubernetes/kubernetes/pull/83906), [@michaelgugino](https://github.com/michaelgugino)) [SIG API Machinery, Apps, Node and Scheduling]

another and become parallel, the blue end _m_ of the Image _mn_ being by
a greater Refraction translated farther from its former place MT, than
the red end _t_ of the other Image _pt_ from the same place MT; which
puts the Proposition past Dispute. And this happens whether the third
Prism DH be placed immediately after the two first, or at a great
distance from them, so that the Light refracted in the two first Prisms