Topology constraints addressed the first point, but up to now pod scheduling was still done without considering that the remaining storage capacity may not be enough to start a new pod. [Storage capacity tracking](https://github.com/kubernetes/enhancements/tree/master/keps/sig-storage/1472-storage-capacity-tracking), a new alpha feature, addresses that by adding an API for a CSI driver to report storage capacity and uses that information in the Kubernetes scheduler when choosing a node for a pod....

### `TensorRT`

*   Issues a warning when the `session_config` parameter for the TF1 converter
    is used or the `rewrite_config_template` field in the TF2 converter
    parameter object is used.

### TPU Enhancements:

*   Adds support for the `beta` parameter of the FTRL optimizer for TPU
    embeddings. Users of other TensorFlow platforms can implement equivalent
    behavior by adjusting the `l2` parameter.