$1.$ What are the solutions used by TensorFlow? $($A$)$ It offers flexible and different dataflow$-$based programming abstractions to adapt applications on heterogeneous hardware platforms; $($B$)$ It does not support multiple concurrent executions on overlapping sub$-$graphs; $($C$)$ Mutable states are owned by individual vertices, and cannot be shared between different executions of the graph; $($D$)$ It cannot use user$-$defined checkpointing for fault tolerance.