*   XLA:GPU reductions are deterministic by default (reductions within
        `jit_compile=True` are now deterministic).
    *   XLA:GPU works with Horovod (OSS contribution by Trent Lo from NVidia)
    *   XLA:CPU and XLA:GPU can compile tf.unique and tf.where when shapes are
        provably correct at compile time.
*   `tf.saved_model.save`:

0008serveftp\u0009servegame\rservehalflife\u0009servehttp\u000bservehumour\u0008serveirc\u0008servemp3\u0008servep2p\u0009servepics\nservequake\u000cservesarcasm\nshopitsite\u0007siiites\nsimple-url\nsimplesite\u0007sinaapp\u0008smushcdn\rspace-to-rent\u0011stackhero-network\u0006stdlib\u0009strapiapp\u000bstreak-link\u000bstreaklinks\u0011streakusercontent\u000cstreamlitapp\u000bstufftoread\u0009subsc-pay\u000ftaveusercontent\ntb-hosting\u000cteaches-yoga\u0008temp-dns\u0009theworkpc\rthingdust...