Tensorflow serving grpc response compression?

Hi,

Is there a way for tensorflow serving to compress the grpc response? I know there’s an option to compress the gprc request, but am wondering if the same option is available for the response.

Thanks.

Hi @chow ,

In essence, while TensorFlow Serving can support compressed gRPC responses, it’s not a built-in feature and requires low-level gRPC configuration changes on the server side.

  1. Compression is handled at the gRPC level for both requests and responses.
  2. Enabling response compression requires server-side configuration changes.
  3. Configuration involves modifying the gRPC C++ implementation used by TensorFlow Serving.
  4. This typically requires changes to the TensorFlow Serving source code and recompilation.

Thank You .