*   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

padding-left: $custom-control-gutter + $custom-control-indicator-size;\n}\n\n.custom-control-inline {\n  display: inline-flex;\n  margin-right: $custom-control-spacer-x;\n}\n\n.custom-control-input {\n  position: absolute;\n  left: 0;\n  z-index: -1; // Put the input behind the label so it doesn't overlay text\n  width: $custom-control-indicator-size;\n  height: ($font-size-base * $line-height-base + $custom-control-indicator-size) / 2;\n  opacity: 0;\n\n  &:checked ~ .custom-control-label::before {\n...