That means that, for example, to apply authentication to all the *path operations* in a router it would end up being done in a different file, instead of keeping related logic together.

*   Added a distributed runtime using GRPC
*   Move skflow to `contrib/learn`
*   Better linear optimizer in `contrib/linear_optimizer`
*   Random forest implementation in `contrib/tensor_forest`
*   CTC loss and decoders in `contrib/ctc`
*   Basic support for `half` data type
*   Better support for loading user ops (see examples in `contrib/`)
*   Allow use of (non-blocking) Eigen threadpool with

  patent or other intellectual property rights of any other entity.
  Each Contributor disclaims any liability to Recipient for claims
  brought by any other entity based on infringement of intellectual
  property rights or otherwise. As a condition to exercising the
  rights and licenses granted hereunder, each Recipient hereby
  assumes sole responsibility to secure any other intellectual