aclnnKlDivTargetBackward

Applicable Products

Product	Supported
Ascend 950PR/Ascend 950DT	×
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	×

Function

Operator function: Performs the target backward computation for the result of the aclnnKlDiv API.
Formula:

gradTarget = \begin{cases} gradOutput * (log(target) - self + 1), ~~~~~~~~~~~~~~~~~~~~~~~~~logTarget=False \\ gradOutput * exp(target) * (target - self +1), ~~~~~~~~~~logTarget=True \end{cases}

Function Prototype

Each operator is divided into a two-phase API. First call the "aclnnKlDivTargetBackwardGetWorkspaceSize" API to obtain the workspace size required for computation and the executor that contains the operator computation process, and then call the "aclnnKlDivTargetBackward" API to perform the computation.

aclnnStatus aclnnKlDivTargetBackwardGetWorkspaceSize(
    const aclTensor* gradOutput,
    const aclTensor* self,
    const aclTensor* target,
    int64_t          reduction,
    bool             logTarget,
    aclTensor*       gradTarget,
    uint64_t*        workspaceSize,
    aclOpExecutor**  executor)

aclnnStatus aclnnKlDivTargetBackward(
    void*          workspace,
    uint64_t       workspaceSize,
    aclOpExecutor* executor,
    aclrtStream    stream)

aclnnKlDivTargetBackwardGetWorkspaceSize

Parameters:

Parameter	Input/Output	Description	Instruction	Data Type	Data Format	Shape	Non-contiguous Tensor
gradOutput (aclTensor*)	Input	Gradient backward input.	The shape must satisfy the broadcast relationship with self.	FLOAT, FLOAT16, BFLOAT16	ND	1-8	√
self (aclTensor*)	Input	Input tensor.	-	FLOAT, FLOAT16, BFLOAT16	ND	1-8	√
target (aclTensor*)	Input.	The ground-truth label.	The shape must satisfy the broadcast relationship with self.	FLOAT, FLOAT16, BFLOAT16	ND	1-8	√
reduction (int64_t)	Input	Specifies the reduction to apply to the output.	Supports 0 (none), 1 (mean), 2 (sum), and 3 (batchmean). 'none' indicates that no reduction is applied. 'mean' indicates that the sum of the output is divided by the number of elements in the output. 'sum' indicates that the output is summed. 'batchmean' indicates that the sum of the output is divided by the batch size.	INT64	-	-	-
logTarget (bool)	Input	Specifies whether to perform log-space transformation on target.	-	BOOL	-	-	√
gradTarget (aclTensor*)	Output	Output loss.	-	Consistent with self.	ND	1-8	√
workspaceSize (uint64_t*)	Output	Returns the workspace size to be allocated on the Device side.	-	-	-	-	-
executor (aclOpExecutor**)	Output	Returns the op executor, which contains the operator computation flow.	-	-	-	-	-

Return value

aclnnStatus: return code. For details, see aclnn Return Code.

The first-phase API performs input parameter validation and returns an error in the following scenarios:

Return Value	Error Code	Description
ACLNN_ERR_PARAM_NULLPTR	161001	The input gradOutput, self, target, and gradTarget are null pointers.
ACLNN_ERR_PARAM_INVALID	161002	The data types of gradOutput, self, target, and gradTarget are not within the supported range.
		The data types of target and gradTarget are inconsistent.
		The shape of gradOutput cannot be broadcast to self or target.
		The shape of target and the shape of self do not satisfy the broadcast relationship.
		The shape of target is different from the shape of gradTarget.
		The dimension of gradOutput, self, target, or gradTarget is greater than 8.

aclnnKlDivTargetBackward

Parameters:

Parameter	Input/Output	Description
workspace	Input	Workspace memory address applied for on the Device side.
workspaceSize	Input	Workspace size applied for on the Device side, obtained from the first-phase API aclnnKlDivTargetBackwardGetWorkspaceSize.
executor	Input	Operator executor that contains the operator computation flow.
stream	Input	Specifies the stream for task execution.

Return value

aclnnStatus: return code. For details, see aclnn Return Code.

Constraints

Deterministic computation:
- aclnnKlDivTargetBackward defaults to a deterministic implementation.

Example

The following provides example code for reference only. For details about compilation and running, see Compile and Run Samples.

#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_kl_div_target_backward.h"

#define CHECK_RET(cond, return_expr) \
  do {                               \
    if (!(cond)) {                   \
      return_expr;                   \
    }                                \
  } while (0)

#define LOG_PRINT(message, ...)     \
  do {                              \
    printf(message, ##__VA_ARGS__); \
  } while (0)

int64_t GetShapeSize(const std::vector<int64_t>& shape) {
  int64_t shapeSize = 1;
  for (auto i : shape) {
    shapeSize *= i;
  }
  return shapeSize;
}

int Init(int32_t deviceId, aclrtStream* stream) {
  // Boilerplate: resource initialization.
  auto ret = aclInit(nullptr);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclInit failed. ERROR: %d\n", ret); return ret);
  ret = aclrtSetDevice(deviceId);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSetDevice failed. ERROR: %d\n", ret); return ret);
  ret = aclrtCreateStream(stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtCreateStream failed. ERROR: %d\n", ret); return ret);
  return 0;
}

template <typename T>
int CreateAclTensor(const std::vector<T>& hostData, const std::vector<int64_t>& shape, void** deviceAddr,
                    aclDataType dataType, aclTensor** tensor) {
  auto size = GetShapeSize(shape) * sizeof(T);
  // Call aclrtMalloc to allocate device memory.
  auto ret = aclrtMalloc(deviceAddr, size, ACL_MEM_MALLOC_HUGE_FIRST);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtMalloc failed. ERROR: %d\n", ret); return ret);

  // Call aclrtMemcpy to copy data from the host to the device memory.
  ret = aclrtMemcpy(*deviceAddr, size, hostData.data(), size, ACL_MEMCPY_HOST_TO_DEVICE);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtMemcpy failed. ERROR: %d\n", ret); return ret);

  // Calculate the strides of a contiguous tensor.
  std::vector<int64_t> strides(shape.size(), 1);
  for (int64_t i = shape.size() - 2; i >= 0; i--) {
    strides[i] = shape[i + 1] * strides[i + 1];
  }

  // Call the aclCreateTensor API to create an aclTensor.
  *tensor = aclCreateTensor(shape.data(), shape.size(), dataType, strides.data(), 0, aclFormat::ACL_FORMAT_ND,
                            shape.data(), shape.size(), *deviceAddr);
  return 0;
}

int main() {
  // 1. (Boilerplate) Initialize the device/stream. See the ACL API manual.
  // Set deviceId based on the actual device.
  int32_t deviceId = 0;
  aclrtStream stream;
  auto ret = Init(deviceId, &stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("Init acl failed. ERROR: %d\n", ret); return ret);

  // 2. Construct the input and output. Customize the construction based on the API interface.
  std::vector<int64_t> gradOutputShape = {2, 2};
  std::vector<int64_t> selfShape = {2, 2};
  std::vector<int64_t> targetShape = {2, 2};
  std::vector<int64_t> gradTargetShape = {2, 2};
  void* gradOutputDeviceAddr = nullptr;
  void* selfDeviceAddr = nullptr;
  void* targetDeviceAddr = nullptr;
  void* gradTargetDeviceAddr = nullptr;
  aclTensor* gradOutput = nullptr;
  aclTensor* self = nullptr;
  aclTensor* target = nullptr;
  aclTensor* gradTarget = nullptr;
  std::vector<float> gradOutputHostData = {2, 3, 5, 8};
  std::vector<float> selfHostData = {2, 3, 5, 8};
  std::vector<float> targetHostData = {2, 3, 5, 8};
  std::vector<float> gradTargetHostData = {2, 3, 5, 8};
  int64_t reduction = 0;
  bool logTarget = false;
  // Create the gradOutput aclTensor.
  ret = CreateAclTensor(gradOutputHostData, gradOutputShape, &gradOutputDeviceAddr, aclDataType::ACL_FLOAT, &gradOutput);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create the self aclTensor.
  ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create the target aclTensor.
  ret = CreateAclTensor(targetHostData, targetShape, &targetDeviceAddr, aclDataType::ACL_FLOAT, &target);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create the gradTarget aclTensor.
  ret = CreateAclTensor(gradTargetHostData, gradTargetShape, &gradTargetDeviceAddr, aclDataType::ACL_FLOAT, &gradTarget);
  CHECK_RET(ret == ACL_SUCCESS, return ret);

  // 3. Call the CANN operator library API.
  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;
  // Call the first-phase API of aclnnKlDivTargetBackward.
  ret = aclnnKlDivTargetBackwardGetWorkspaceSize(gradOutput, self, target, reduction, logTarget, gradTarget, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnKlDivTargetBackwardGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
  // Apply for device memory based on the workspaceSize calculated by the first-phase API.
  void* workspaceAddr = nullptr;
  if (workspaceSize > 0) {
    ret = aclrtMalloc(&workspaceAddr, workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("allocate workspace failed. ERROR: %d\n", ret); return ret);
  }
  // Call the second-phase API of aclnnKlDivTargetBackward.
  ret = aclnnKlDivTargetBackward(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnKlDivTargetBackward failed. ERROR: %d\n", ret); return ret);

  // 4. (Boilerplate) Synchronize and wait for task execution to complete.
  ret = aclrtSynchronizeStream(stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSynchronizeStream failed. ERROR: %d\n", ret); return ret);

  // 5. Obtain the output value, copy the result from the device memory to the host, and modify based on the specific API definition.
  auto size = GetShapeSize(gradTargetShape);
  std::vector<float> resultData(size, 0);
  ret = aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]),
                    gradTargetDeviceAddr, size * sizeof(resultData[0]), ACL_MEMCPY_DEVICE_TO_HOST);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("copy resultData from device to host failed. ERROR: %d\n", ret);
            return ret);
  for (int64_t i = 0; i < size; i++) {
    LOG_PRINT("resultData[%ld] is: %f\n", i, resultData[i]);
  }

  // 6. Release the aclTensor. Modify based on the specific API definition.
  aclDestroyTensor(gradOutput);
  aclDestroyTensor(self);
  aclDestroyTensor(target);
  aclDestroyTensor(gradTarget);

  // 7. Release the device resources. Modify based on the specific API definition.
  aclrtFree(gradOutputDeviceAddr);
  aclrtFree(selfDeviceAddr);
  aclrtFree(targetDeviceAddr);
  aclrtFree(gradTargetDeviceAddr);
  if (workspaceSize > 0) {
    aclrtFree(workspaceAddr);
  }
  aclrtDestroyStream(stream);
  aclrtResetDevice(deviceId);
  aclFinalize();
  return 0;
}