aclnnBinaryCrossEntropyBackward

📄 View Source Code

Applicable Products

Function

• Description: Computes the gradient value of binary cross-entropy backward propagation.

• Formula:

  The formula for binary cross-entropy is as follows:

  $$y_i = - weight_i \cdot (target_i \cdot ln(x_i) + (1 - target_i) \cdot ln(1- x_i))$$

  Where $x$ represents the output of the previous network layer, i.e., the forward prediction value; $target$ represents the label value of the sample. Find the partial derivative of the binary cross-entropy with respect to $x$:

  $$\begin{aligned} \frac {\partial y}{\partial x} &= -weight \cdot \frac {\partial (target_i \cdot ln(x_i) + (1-target_i) \cdot ln(1-x_i))}{\partial x} \\ &= -weight \cdot (\frac {\partial (target \cdot ln(x))}{\partial x} + \frac {\partial ((1-target) \cdot ln(1-x))}{\partial x}) \\ &= -weight \cdot (\frac {target}{x} - \frac {(1-target)}{1-x}) \\ &= -weight \cdot \frac {target(1-x)-x(1-target)}{x(1-x)} \\ &= -weight \cdot \frac {target-x}{x(1-x)} \\ &= weight \cdot \frac {x-target}{x(1-x)} \end{aligned}$$ $$out = grad\_output \cdot weight \cdot \frac {x-target}{x(1-x)} \\ out = mean(grad\_output) \ if \ reduction = mean$$

Function Prototype

Each operator is divided into a two-phase API. You must first call the "aclnnBinaryCrossEntropyBackwardGetWorkspaceSize" API to obtain the input parameters and calculate the required workspace size based on the process, and then call the "aclnnBinaryCrossEntropyBackward" API to perform the computation.

aclnnStatus aclnnBinaryCrossEntropyBackwardGetWorkspaceSize(
    const aclTensor *gradOutput,
    const aclTensor *self,
    const aclTensor *target,
    const aclTensor *weightOptional,
    int64_t          reduction,
    aclTensor       *out,
    uint64_t        *workspaceSize,
    aclOpExecutor  **executor)

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

aclnnBinaryCrossEntropyBackwardGetWorkspaceSize

• Parameters:

  Parameter	Input/Output	Description	Instruction	Data Type	Data Format	Dimension (Shape)	Non-Contiguous Tensor
  gradOutput (aclTensor*)	Input	Gradient value from the previous step of network backward propagation. The data type needs to be converted to the promotion type together with other parameters.	The shape can be broadcast to the shape of self.	FLOAT, FLOAT16, BFLOAT16	ND	0-8	√
  self (aclTensor*)	Input	Computation result of the previous layer in the forward pass of the network.	The data type needs to be converted to the promotion type together with other parameters.	FLOAT, FLOAT16, BFLOAT16	ND	0-8	√
  target (aclTensor*)	Input	Label value of the sample. The data type needs to be converted together with other parameters to the promotion type.	The shape can be broadcast to the shape of self.	FLOAT, FLOAT16, BFLOAT16	Must be consistent with self.	0-8	√
  weightOptional (aclTensor*)	Input.	Weight of the result.	The shape can be broadcast to the shape of self. When weightOptional is empty, a tensor of all ones needs to be created with the shape of self.	FLOAT, FLOAT16, BFLOAT16	Must be consistent with self.	0-8	√
  reduction (int64_t)	Input	Specifies the reduce operation to perform on the gradient result of binary cross-entropy backward computation.	Only 0 (none), 1 (mean), and 2 (sum) are supported: 0: No operation is performed. 1: Computes the mean of the result. 2: Computes the sum of the result.	INT64	-	-	-
  out (aclTensor*)	Output	Stores the gradient computation result, with the same shape as self.	-	FLOAT, FLOAT16, BFLOAT16	-	0-8	√
  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.	-	-	-	-	-

  • Atlas inference products, Atlas training products: The BFLOAT16 data type is not supported.

• 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 passed gradOutput, self, target, or out is a null pointer.
  ACLNN_ERR_PARAM_INVALID	161002	The data type of gradOutput, self, target, or weightOptional is not within the supported range.
  		The scenario where self is a scalar and gradOutput is a non-scalar is not supported.
  		The shape of gradOutput, target, and weightOptional cannot be broadcast to the shape of self.
  		The value of reduction is not one of 0, 1, or 2.
  		The shape dimensions of gradOutput, self, target, and weightOptional are greater than 8.

aclnnBinaryCrossEntropyBackward

• Parameters:

  Parameter	Input/Output	Description
  workspace	Input	Specifies the workspace memory address applied for on the device side.
  workspaceSize	Input	Specifies the workspace size applied for on the device side, which is obtained by the first-phase API aclnnBinaryCrossEntropyBackwardGetWorkspaceSize.
  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 Code.

Constraints

• Deterministic computation:

  • aclnnBinaryCrossEntropyBackward 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_binary_cross_entropy_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 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 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);

  // Copy data from the host side to the device side by calling aclrtMemcpy.
  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];
  }

  // Create an aclTensor by calling the aclCreateTensor API.
  *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);
  // Handle the check as needed.
  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.
  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> weightShape = {2, 2};
  std::vector<int64_t> outShape = {2, 2};
  void* gradOutputDeviceAddr = nullptr;
  void* selfDeviceAddr = nullptr;
  void* targetDeviceAddr = nullptr;
  void* weightDeviceAddr = nullptr;
  void* outDeviceAddr = nullptr;
  aclTensor* gradOutput = nullptr;
  aclTensor* self = nullptr;
  aclTensor* target = nullptr;
  aclTensor* weight = nullptr;
  aclTensor* out = nullptr;
  std::vector<float> gradOutputHostData = {0.1, 0.1, 0.1, 0.1};
  std::vector<float> selfHostData = {0.3, 0.3, 0.3, 0.3};
  std::vector<float> targetHostData = {0.5, 0.5, 0.5, 0.5};
  std::vector<float> weightHostData = {1, 1, 1, 1};
  std::vector<float> outHostData = {0, 0, 0, 0};
  int64_t reduction = 0;
  // Create gradOutput aclTensor.
  ret = CreateAclTensor(gradOutputHostData, gradOutputShape, &gradOutputDeviceAddr, aclDataType::ACL_FLOAT, &gradOutput);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create self aclTensor.
  ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create target aclTensor.
  ret = CreateAclTensor(targetHostData, targetShape, &targetDeviceAddr, aclDataType::ACL_FLOAT, &target);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create the weight aclTensor.
  ret = CreateAclTensor(weightHostData, weightShape, &weightDeviceAddr, aclDataType::ACL_FLOAT, &weight);
  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);

  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;

  // Example of aclnnBinaryCrossEntropyBackward API call
  // 3. Call the CANN operator library API. Replace with the specific API name.
  // Call the first-phase API of aclnnBinaryCrossEntropyBackward.
  ret = aclnnBinaryCrossEntropyBackwardGetWorkspaceSize(gradOutput, self, target, weight, reduction, out, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnBinaryCrossEntropyBackwardGetWorkspaceSize 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 aclnnBinaryCrossEntropyBackward.
  ret = aclnnBinaryCrossEntropyBackward(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnBinaryCrossEntropyBackward failed. ERROR: %d\n", ret); return ret);

  // 4. (boilerplate) Synchronously 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 and copy the result from the Device side memory to the Host side. Modify 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(resultData[0]), 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(gradOutput);
  aclDestroyTensor(self);
  aclDestroyTensor(target);
  aclDestroyTensor(weight);
  aclDestroyTensor(out);

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