aclnnSoftmaxBackward

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: Performs backward propagation of softmax.
Formula: For the derivative of the Softmax function, the following formula can be used:

The relationship between out (input gradient), gradOutput (output gradient of the previous layer), and output (forward output of Softmax) can be expressed as follows:
$out = gradOutput \cdot output - sum(gradOutput \cdot output)\cdot output$

Function Prototype

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

aclnnStatus aclnnSoftmaxBackwardGetWorkspaceSize(
  const aclTensor *gradOutput,
  const aclTensor *output,
  int64_t          dim,
  aclTensor*       out,
  uint64_t        *workspaceSize,
  aclOpExecutor  **executor)

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

aclnnSoftmaxBackwardGetWorkspaceSize

Parameters:

Parameter	Input/Output	Description	Instruction	Data Type	Data Format	Dimension (Shape)	Non-contiguous Tensor
gradOutput (aclTensor*)	Input	Specifies the gradient value of backward propagation, that is, the output gradient of the previous layer. Corresponds to gradOutput in the formula.	Supports empty tensors.	FLOAT16, FLOAT32, BFLOAT16	ND	0-8	√
output (aclTensor*)	Input	Output value of the Softmax function, i.e., output in the formula.	Empty tensors are supported. The shape and data type must be consistent with gradOutput	FLOAT16, FLOAT32, BFLOAT16	ND	0-8	√
dim (int64_t)	Input	Specifies the dimension along which the Softmax function is applied.	-	-	-	-	√
out (aclTensor*)	Output	The output of the function is the gradient value of the input.	Result after differentiating the input. Supports multiple data types and does not require additional space allocation. Other data types (FP64, COMPLEX64, etc.) are supported through the automatic cast capability but require additional space allocation. The shape must be consistent with gradOutput	FLOAT16, FLOAT32, BFLOAT16	ND	0-8	√
workspaceSize (uint64_t*)	Output	Returns the workspace size to be requested on the Device side.	-	-	-	-	-
executor (aclOpExecutor**)	Output	Returns the operator 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 Code	Error Code	Description
ACLNN_ERR_PARAM_NULLPTR	161001	The input gradOutput, output, or out is a null pointer.
ACLNN_ERR_PARAM_INVALID	161002	The dim value is out of the dimension range of gradOutput
		gradOutput is not of a supported data type.
		output is not of a supported data type.
		The shapes of gradOutput, output, and out are inconsistent.

aclnnSoftmaxBackward

Parameters:

Parameter	Input/Output	Description
workspace	Input	Specifies the workspace memory address applied for on the device.
workspaceSize	Input	Specifies the workspace size applied for on the device, which is obtained by the first-phase API aclnnSoftmaxBackwardGetWorkspaceSize.
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:
- aclnnSoftmaxBackward 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_softmax_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 apply for device-side 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 the host-side data to the device-side memory 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. Refer to the ACL API manual.
  // Set the deviceId based on your 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.
  std::vector<int64_t> gradOutputShape = {4, 2};
  std::vector<int64_t> outputShape = {4, 2};
  std::vector<int64_t> outShape = {4, 2};
  void* gradOutputDeviceAddr = nullptr;
  void* outputDeviceAddr = nullptr;
  void* outDeviceAddr = nullptr;
  aclTensor* gradOutput = nullptr;
  aclTensor* output = nullptr;
  aclTensor* out = nullptr;
  std::vector<float> gradOutputHostData = {0, 1, 2, 3, 4, 5, 6, 7};
  std::vector<float> outputHostData = {1, 1, 1, 2, 2, 2, 3, 3};
  std::vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0};
  // Create the gradOutput aclTensor.
  ret = CreateAclTensor(gradOutputHostData, gradOutputShape, &gradOutputDeviceAddr, aclDataType::ACL_FLOAT, &gradOutput);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create the output aclTensor.
  ret = CreateAclTensor(outputHostData, outputShape, &outputDeviceAddr, aclDataType::ACL_FLOAT, &output);
  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. Replace with the specific API name.
  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;
  int64_t dim = 0;
  // Call the first-phase API of aclnnSoftmaxBackward.
  ret = aclnnSoftmaxBackwardGetWorkspaceSize(gradOutput, output, dim, out, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnSoftmaxBackwardGetWorkspaceSize 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 aclnnSoftmaxBackward.
  ret = aclnnSoftmaxBackward(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnSoftmaxBackward 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(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(output);
  aclDestroyTensor(out);

  // 7. Release device resources.
  aclrtFree(gradOutputDeviceAddr);
  aclrtFree(outputDeviceAddr);
  aclrtFree(outDeviceAddr);
  if (workspaceSize > 0) {
    aclrtFree(workspaceAddr);
  }
  aclrtDestroyStream(stream);
  aclrtResetDevice(deviceId);
  aclFinalize();

  return 0;
}