LoadFromData

Applicability

Product	Supported or Not
Atlas A3 training products/Atlas A3 inference products	√
Atlas A2 training products/Atlas A2 inference products	√
Atlas 200I/500 A2 inference products	√
Atlas inference products	√
Atlas training products	√

Header File

#include <graph/kernel_launch_info.h>

Function Usage

Deserializes a serialized task. This API is used to deserialize the task constructed by the framework in the GenerateTask function and modify the task parameters.

Prototype

static KernelLaunchInfo LoadFromData(const gert::ExeResGenerationContext *context, const std::vector<uint8_t> &data)

Parameters

Parameter	Input/Output	Description
context	Input	Input parameter of the GenerateTask function, which stores the basic information about the operator.
data	Input	Result after KernelLaunchInfo is serialized.

Returns

Deserialization result of the input parameter data.

Constraints

None

Examples

graphStatus Mc2GenTaskCallback(const gert::ExeResGenerationContext *context,
    std::vector<std::vector<uint8_t>> &tasks) {
   ....
  // Change ArgsFormat of the original AI Core task.
  auto aicore_task = KernelLaunchInfo::LoadFromData(context, tasks.back());
  // Obtain argsformat from the task.
  auto aicore_args_format_str = aicore_task.GetArgsFormat();
  auto aicore_args_format = ArgsFormatSerializer::Deserialize(aicore_args_format_str);
  size_t i = 0UL;
  // Find the position of the first input address.
  for (; i < aicore_args_format.size(); i++) {
    if (aicore_args_format[i].GetType() == ArgDescType::kIrInput ||
        aicore_args_format[i].GetType() == ArgDescType::kInputInstance) {
      break;
    }
  }
  // Insert a communication address before the input address.
  aicore_args_format.insert(aicore_args_format.begin() + i, ArgDescInfo::CreateHiddenInput(HiddenInputSubType::kHcom));
  // Serialize the modified argsformat and set it back to the task.
  aicore_task.SetArgsFormat(ArgsFormatSerializer::Serialize(aicore_args_format).GetString());
  // argsformat is reserialized and set in the task.
  tasks.back() = aicore_task.Serialize();
  return SUCCESS;
}

Parent topic: KernelLaunchInfo