Class ExpandInlineFunctionsOptions

tensorflow/core/common_runtime/function.h:326–334

There are three types of function calls that could be invoked during Tensorflow graph execution*: 1) Native function call (node.type_string() is the function name). These functions are always executed on a single-device, which is the device of the function call node. 2) Multi-device function calls (PartitionedCall or StatefulPartitionedCall ops) can execute on multiple devices and accept DT_RESO

// 2) tf.data pipelines - also execute functions directly via function library
//    runtime with custom executors.
struct ExpandInlineFunctionsOptions {
  ExpandInlineFunctionsOptions() : native_options(), multi_device_options() {
    using OutputControlSrc = InlineFunctionBodyOptions::OutputControlSource;
    multi_device_options.output_control_src = OutputControlSrc::kControlOutputs;
  }

  InlineFunctionBodyOptions native_options;
  InlineFunctionBodyOptions multi_device_options;
};

// WARNING(ezhulenev): PLEASE DO NOT USE THIS FUNCTION. This is a temporary
// workaround that will be enabled only during the function inlining unification