aclnnKlDiv

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

Description: Computes the KL divergence.
Formula:
- Defines loss_pointwise to store intermediate results.
  $loss\_pointwise_i=\begin{cases} NaN & \text{ if }&logTarget=false \text{ and } target_i <= 0, \\ target_i * \left ( \log{(target_i)}- self_i \right ) & \text{ if }& logTarget=false, \\ \exp^ {target_i} * \left ( target_i- self_i \right ) & \text{ else. } \end{cases}$
- The formula for out is:
  $out=\begin{cases} \bar{loss\_pointwise} & \text{ if }& reduction= 1, \\ \sum loss\_pointwise & \text{ elif }& reduction= 2,\\ \frac{\sum loss\_pointwise}{self.size(0)} & \text{ elif }& reduction= 3,\\ loss\_pointwise & \text{ else. } \end{cases}$

Function Prototype

Each operator is divided into a two-phase API. The "aclnnKlDivGetWorkspaceSize" API must be called first to obtain the workspace size required for computation and the executor that encapsulates the operator computation flow. Then, the "aclnnKlDiv" API is called to perform the computation.

aclnnStatus aclnnKlDivGetWorkspaceSize(
  const aclTensor*          self,
  const aclTensor*          target,
  int64_t                   reduction,
  bool                      logTarget,
  aclTensor*                out,
  uint64_t*                 workspaceSize,
  aclOpExecutor**           executor)

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

aclnnKlDivGetWorkspaceSize

Parameters:

Parameter	Input/Output	Description	Instruction	Data Type	Data Format	Dimension	Non-contiguous Tensor
self (aclTensor*)	Input	self in the formula.	The data type and target must satisfy the data type deduction rule (see deduction relationship). The shape must satisfy the broadcast relationship with target.	FLOAT, FLOAT16, BFLOAT16	ND	0-8	√
target (aclTensor*)	Input	Target in the formula.	The data type must comply with the data type deduction rule of self (see deduction relationship). The shape must satisfy the broadcast relationship with self.	FLOAT, FLOAT16, BFLOAT16	ND	0-8	√
reduction (int64_t)	Input	Reduction in the formula.	0: none, indicating no reduction operation. 1: mean, indicating the mean of loss_pointwise. 2: sum, indicating the sum of loss_pointwise. 3: batchmean, indicating the average loss of the batch, consistent with the mathematical definition of KL divergence.	-	-	-	-
logTarget (bool)	Input	Specifies whether the input target data has been log-transformed.	-	-	-	-	-
out (aclTensor*)	Output	out in the formula.	The data type meets the data type deduction rule with self (see deduction relationship). When `reduction` is 0, the shape must satisfy the broadcast relationship with self and target. When `reduction` is not 0, the shape is fixed to (1,).	FLOAT, FLOAT16, BFLOAT16	ND	0-8	√
workspaceSize (uint64_t*)	Output	Returns the workspace size to be applied for on the Device side.	-	-	-	-	-
executor (aclOpExecutor**)	Output	Returns the op executor, which contains the operator computation flow.	-	-	-	-	-

Atlas inference products, Atlas training products: BFLOAT16 is not supported.

Return value

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

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 passed self, target, or out is a null pointer.
ACLNN_ERR_PARAM_INVALID	161002	The data type of self, target, or out is not within the supported range.
		The shape of self cannot be broadcast with target.
		self or target cannot be converted to the data type of out.
		The shape dim of self or target is greater than 8.

aclnnKlDiv

Parameters:

Parameter	Input/Output	Description
workspace	Input	Address of the workspace memory applied on the Device side.
workspaceSize	Input	Size of the workspace applied on the Device side, obtained by the first-phase API aclnnKlDivGetWorkspaceSize.
executor	Input	Operator executor, which contains the operator computation flow.
stream	Input	Specifies the stream for task execution.

Return value

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

Constraints

Deterministic computation:
- aclnnKlDiv 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.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 shape_size = 1;
  for (auto i : shape) {
    shape_size *= i;
  }
  return shape_size;
}

int Init(int32_t deviceId, aclrtStream* stream) {
  // Boilerplate for 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 memory on the device side.
  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 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.
  // Fill in the deviceId based on your actual device.
  int32_t deviceId = 0;
  aclrtStream stream;
  auto ret = Init(deviceId, &stream);
  // Handle the check based on your needs.
  CHECK_RET(ret == 0, LOG_PRINT("Init acl failed. ERROR: %d\n", ret); return ret);
  // 2. Construct the input and output. Customize the construction based on the API.
  std::vector<int64_t> selfShape = {4, 2};
  std::vector<int64_t> targetShape = {4, 2};
  std::vector<int64_t> outShape = {1};
  void* selfDeviceAddr = nullptr;
  void* targetDeviceAddr = nullptr;
  void* outDeviceAddr = nullptr;
  aclTensor* self = nullptr;
  aclTensor* target = nullptr;
  aclTensor* out = nullptr;
  std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7};
  std::vector<float> targetHostData = {1, 1, 1, 2, 2, 2, 3, 3};
  std::vector<float> outHostData = {1};
  int64_t reduction = 1;
  bool log_target = false;
  // 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 out aclTensor.
  ret = CreateAclTensor(outHostData, outShape, &outDeviceAddr, aclDataType::ACL_FLOAT, &out);
  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 aclnnKlDiv.
  ret = aclnnKlDivGetWorkspaceSize(self, target, reduction, log_target, out, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnKlDivGetWorkspaceSize 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 aclnnKlDiv.
  ret = aclnnKlDiv(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnKlDiv failed. ERROR: %d\n", ret); return ret);
  // 4. (Boilerplate) Synchronously wait for the 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 and copy the result from the device memory to the host memory. Modify this based on the specific API definition.
  auto size = GetShapeSize(outShape);
  std::vector<float> resultData(size, 0);
  ret = aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]), outDeviceAddr, size * sizeof(float),
                    ACL_MEMCPY_DEVICE_TO_HOST);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("copy result from device to host failed. ERROR: %d\n", ret); return ret);
  for (int64_t i = 0; i < size; i++) {
    LOG_PRINT("result[%ld] is: %f\n", i, resultData[i]);
  }

  // 6. Release aclTensor and aclScalar. Modify based on the specific API definition.
  aclDestroyTensor(self);
  aclDestroyTensor(target);
  aclDestroyTensor(out);

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