aclnnForeachNonFiniteCheckAndUnscale

📄 View Source Code

Applicable Products

Function

• Description: Traverses all tensors in scaledGrads, checks whether Inf or NaN exists. If so, sets foundInf to 1.0; otherwise, foundInf remains unchanged. Then unscales all tensors in scaledGrads.

• Formula:

  $$foundInf = \begin{cases}1.0, & \text{when Inf} \in \text{scaledGrads or NaN} \in \text{scaledGrads},\\ \text{foundInf}, &\text{otherwise}. \end{cases}$$ $$scaledGrads_i = {scaledGrads}_{i}*{invScale}.$$

Function Prototype

Each operator is divided into a two-phase API. First call the "aclnnForeachNonFiniteCheckAndUnscaleGetWorkspaceSize" API to obtain the input parameters and calculate the required workspace size based on the computation process, and then call the "aclnnForeachNonFiniteCheckAndUnscale" API to perform the computation.

aclnnStatus aclnnForeachNonFiniteCheckAndUnscaleGetWorkspaceSize(
  const aclTensorList *scaledGrads,
  const aclTensor     *foundInf,
  const aclTensor     *invScale,
  uint64_t            *workspaceSize,
  aclOpExecutor      **executor)

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

aclnnForeachNonFiniteCheckAndUnscaleGetWorkspaceSize

• Parameters

  Parameter	Input/Output	Description	Instruction	Data Type	Data Format	Dimension (Shape)	Non-contiguous Tensor
  scaledGrads (aclTensorList*)	Input/Output	Represents the input and output tensor list for the unscaling computation, corresponding to `scaledGrads` in the formula.	Null tensors are not supported. All tensors in this parameter must have the same data type. The maximum supported length is 256.	FLOAT32, FLOAT16, BFLOAT16	ND	0-8	√
  foundInf (aclTensor*)	Input/Output	Indicates the tensor used to mark whether Inf or -Inf exists in the input scaledGrads, corresponding to `foundInf` in the formula.	Null tensor not supported. Contains only one element.	FLOAT32	ND	0-8	√
  invScale (aclTensor*)	Input	Tensor for inverse scaling computation, corresponding to `invScale` in the formula.	Null not supported. Contains only one element.	FLOAT32	ND	0-8	√
  workspaceSize (uint64_t*)	Output	Returns the workspace size to be applied 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 Code	Error Code	Description
  ACLNN_ERR_PARAM_NULLPTR	161001	The passed scaledGrads, foundInf, or invScale is a null pointer.
  ACLNN_ERR_PARAM_INVALID	161002	The data type of scaledGrads, foundInf, or invScale is not within the supported range.
  ACLNN_ERR_INNER_TILING_ERROR	561002	The length of scaledGrads exceeds the limit.
  		The data types of tensors in scaledGrads are not the same.
  		The number of elements in foundInf or invScale is not 1.

aclnnForeachNonFiniteCheckAndUnscale

• Parameters

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

• Return Value

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

Constraints

• Deterministic computation:

  • aclnnForeachNonFiniteCheckAndUnscale 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_foreach_non_finite_check_and_unscale.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 data from the host side to the device memory using 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);
  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> selfShape1 = {2, 3};
  std::vector<int64_t> selfShape2 = {1, 3};
  std::vector<int64_t> foundInfShape = {1};
  std::vector<int64_t> invScaleShape = {1};
  void* input1DeviceAddr = nullptr;
  void* input2DeviceAddr = nullptr;
  void* foundInfDeviceAddr = nullptr;
  void* invScaleDeviceAddr = nullptr;
  aclTensor* input1 = nullptr;
  aclTensor* input2 = nullptr;
  aclTensor* foundInf = nullptr;
  aclTensor* invScale = nullptr;
  std::vector<float> input1HostData = {1, 2, 3, 4, 5, 6};
  std::vector<float> input2HostData = {7, 8, 9};
  std::vector<float> foundInfHostData = {-1.0f};
  std::vector<float> invScaleHostData = {3.1f};
  // Create the input1 aclTensor.
  ret = CreateAclTensor(input1HostData, selfShape1, &input1DeviceAddr, aclDataType::ACL_FLOAT, &input1);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create the input2 aclTensor.
  ret = CreateAclTensor(input2HostData, selfShape2, &input2DeviceAddr, aclDataType::ACL_FLOAT, &input2);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create the foundInf aclTensor.
  ret = CreateAclTensor(foundInfHostData, foundInfShape, &foundInfDeviceAddr, aclDataType::ACL_FLOAT, &foundInf);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create the invScale aclTensor.
  ret = CreateAclTensor(invScaleHostData, invScaleShape, &invScaleDeviceAddr, aclDataType::ACL_FLOAT, &invScale);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  std::vector<aclTensor*> tempInput{input1, input2};
  aclTensorList* tensorListInput = aclCreateTensorList(tempInput.data(), tempInput.size());

  // 3. Call the CANN operator library API. Replace with the specific API name.
  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;
  // Call the first-phase API of aclnnForeachNonFiniteCheckAndUnscale.
  ret = aclnnForeachNonFiniteCheckAndUnscaleGetWorkspaceSize(tensorListInput, foundInf, invScale, &workspaceSize,
                                                             &executor);
  CHECK_RET(ret == ACL_SUCCESS,
            LOG_PRINT("aclnnForeachNonFiniteCheckAndUnscaleGetWorkspaceSize 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 aclnnForeachNonFiniteCheckAndUnscale.
  ret = aclnnForeachNonFiniteCheckAndUnscale(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnForeachNonFiniteCheckAndUnscale 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 and copy the result from the device memory to the host memory. Modify based on the specific API definition.
  auto size = GetShapeSize(selfShape1);
  std::vector<float> self1Data(size, 0);
  ret = aclrtMemcpy(self1Data.data(), self1Data.size() * sizeof(self1Data[0]), input1DeviceAddr,
                    size * sizeof(self1Data[0]), ACL_MEMCPY_DEVICE_TO_HOST);
  CHECK_RET(ret == ACL_SUCCESS,
            LOG_PRINT("copy self1 result from device to host failed. ERROR: %d\n", ret); return ret);
  for (int64_t i = 0; i < size; i++) {
    LOG_PRINT("self1 result[%ld] is: %f\n", i, self1Data[i]);
  }

  size = GetShapeSize(selfShape2);
  std::vector<float> self2Data(size, 0);
  ret = aclrtMemcpy(self2Data.data(), self2Data.size() * sizeof(self2Data[0]), input2DeviceAddr,
                    size * sizeof(self2Data[0]), ACL_MEMCPY_DEVICE_TO_HOST);
  CHECK_RET(ret == ACL_SUCCESS,
            LOG_PRINT("copy self2 result from device to host failed. ERROR: %d\n", ret); return ret);
  for (int64_t i = 0; i < size; i++) {
    LOG_PRINT("self2 result[%ld] is: %f\n", i, self2Data[i]);
  }

  size = GetShapeSize(foundInfShape);
  std::vector<float> foundInfData(size, 0);
  ret = aclrtMemcpy(foundInfData.data(), foundInfData.size() * sizeof(foundInfData[0]), foundInfDeviceAddr,
                    size * sizeof(foundInfData[0]), ACL_MEMCPY_DEVICE_TO_HOST);
  CHECK_RET(ret == ACL_SUCCESS,
            LOG_PRINT("copy foundInf result from device to host failed. ERROR: %d\n", ret); return ret);
  LOG_PRINT("foundInf result is: %f\n", foundInfData[0]);

  // 6. Release aclTensor. Modify based on the specific API definition.
  aclDestroyTensorList(tensorListInput);
  aclDestroyTensor(foundInf);
  aclDestroyTensor(invScale);

  // 7. Release device resources. Modify based on the specific API definition.
  aclrtFree(input1DeviceAddr);
  aclrtFree(input2DeviceAddr);
  aclrtFree(foundInfDeviceAddr);
  aclrtFree(invScaleDeviceAddr);
  if (workspaceSize > 0) {
    aclrtFree(workspaceAddr);
  }
  aclrtDestroyStream(stream);
  aclrtResetDevice(deviceId);
  aclFinalize();
  return 0;
}