aclnnBernoulliTensor&aclnnInplaceBernoulliTensor

Supported Products

Product	Supported
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

Description: Draws binary random numbers (0 or 1) from a Bernoulli distribution. prob indicates the probability of generating 1. The input tensor is used to specify the shape.
Formula:
$out∼Bernoulli(prob)$
When aclnnBernoulliTensor is used, prob and out in the formula correspond to those in the prototype of the first-phase API. When aclnnInplaceBernoulliTensor is used, prob and out in the formula correspond to prob and selfRef respectively in the prototype of the first-phase API.

Prototype

aclnnBernoulliTensor and aclnnInplaceBernoulliTensor implement the same function in different ways. Select a proper operator based on your requirements.
- aclnnBernoulliTensor: An output tensor object needs to be created to store the computation result.
- aclnnInplaceBernoulliTensor: No output tensor object needs to be created, and the computation result is written in place to the input tensor's memory.
Each operator has two-phase API calls. First,·aclnnBernoulliTensorGetWorkspaceSize or aclnnInplaceBernoulliTensorGetWorkspaceSize is called to obtain the workspace size required for computation and the executor covering the operator computation process. Then, aclnnBernoulliTensor or aclnnInplaceBernoulliTensor is called to perform computation.
- aclnnStatus aclnnBernoulliTensorGetWorkspaceSize(const aclTensor* self, const aclTensor* prob, int64_t seed, int64_t offset, aclTensor* out, uint64_t* workspaceSize, aclOpExecutor** executor)
- aclnnStatus aclnnBernoulliTensor(void* workspace, uint64_t workspaceSize, aclOpExecutor* executor, aclrtStream stream)
- aclnnStatus aclnnInplaceBernoulliTensorGetWorkspaceSize(const aclTensor* selfRef, const aclTensor* prob, int64_t seed, int64_t offset, uint64_t* workspaceSize, aclOpExecutor** executor)
- aclnnStatus aclnnInplaceBernoulliTensor(void* workspace, uint64_t workspaceSize, aclOpExecutor* executor, aclrtStream stream)

aclnnBernoulliTensorGetWorkspaceSize

Parameters:
- self (aclTensor*, computation input): which specifies the shape of out and is an aclTensor on the device. The shape supports 0 to 8 dimensions, and must be the same as that of out. It supports non-contiguous tensors. The data format can be ND.
  - Atlas training products: The data type can be FLOAT16, FLOAT, or DOUBLE.
  - Atlas A2 training products/Atlas A2 inference products and Atlas A3 training products/Atlas A3 inference products: The data type can be FLOAT16, FLOAT, DOUBLE, or BFLOAT16.
- prob (aclTensor*, computation input): prob in the formula, aclTensor on the device, which must meet the condition: 0 ≤ prob ≤ 1. The shape supports 0 to 8 dimensions. It supports non-contiguous tensors. The data format must be the same as that of self.
  - Atlas training products: The data type can be FLOAT16, FLOAT, or DOUBLE.
  - Atlas A2 training products/Atlas A2 inference products and Atlas A3 training products/Atlas A3 inference products: The data type can be FLOAT16, FLOAT, DOUBLE, or BFLOAT16.
- seed (int64_t, computation input): integer on the host, used to set the seed of the random number generator.
- offset (int64_t, computation input): integer on the host, used to set the random number offset.
- out (aclTensor*, computation output): out in the formula, aclTensor on the device. The shape supports 0 to 8 dimensions and must be the same as that of self. The data type must be the same as that of self. It supports non-contiguous tensors. The data format can be ND.
  - Atlas training products: The data type can be FLOAT16, FLOAT, or DOUBLE.
  - Atlas A2 training products/Atlas A2 inference products and Atlas A3 training products/Atlas A3 inference products: The data type can be FLOAT16, FLOAT, DOUBLE, or BFLOAT16.
- workspaceSize (uint64_t*, output): size of the workspace to be allocated on the device.
- executor (aclOpExecutor**, output): operator executor, covering the operator computation process.

Returns:

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

The first-phase API implements input parameter verification. The following errors may be thrown:
161001 (ACLNN_ERR_PARAM_NULLPTR): 1. The input self, prob, or out is a null pointer.
161002 (ACLNN_ERR_PARAM_INVALID): 1. The data type or format of self, prob, or out is not supported.
                                      2. The data types of self and out are inconsistent.
                                      3. self, prob, or out has more than 8 dimensions.
                                      4. The shapes of self and out are inconsistent.

aclnnBernoulliTensor

Parameters:
- workspace (void*, input): address of the workspace to be allocated on the device.
- workspaceSize (uint64_t, input): size of the workspace to be allocated on the device, which is obtained by calling aclnnBernoulliTensorGetWorkspaceSize.
- executor (aclOpExecutor*, input): operator executor, covering the operator computation process.
- stream (aclrtStream, input): stream for executing the task.
Returns:

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

aclnnInplaceBernoulliTensorGetWorkspaceSize

Parameters:
- selfRef (aclTensor*, computation input/output): out in the formula, aclTensor on the device. The shape supports 0 to 8 dimensions. It supports non-contiguous tensors. The data format can be ND.
  - Atlas training products: The data type can be FLOAT16, FLOAT, DOUBLE, UINT8, INT8, INT16, INT32, INT64, or BOOL.
  - Atlas A2 training products/Atlas A2 inference products and Atlas A3 training products/Atlas A3 inference products: The data type can be FLOAT16, FLOAT, DOUBLE, UINT8, INT8, INT16, INT32, INT64, BOOL, or BFLOAT16.
- prob (aclTensor*, computation input): prob in the formula, aclTensor on the device, which must meet the condition: 0 ≤ prob ≤ 1. The shape supports 0 to 8 dimensions. It supports non-contiguous tensors. The data format can be ND.
  - Atlas training products: The data type can be FLOAT16, FLOAT, or DOUBLE.
  - Atlas A2 training products/Atlas A2 inference products and Atlas A3 training products/Atlas A3 inference products: The data type can be FLOAT16, FLOAT, DOUBLE, or BFLOAT16.
- seed (int64_t, computation input): integer on the host, used to set the seed of the random number generator.
- offset (int64_t, computation input): integer on the host, used to set the random number offset.
- workspaceSize (uint64_t*, output): size of the workspace to be allocated on the device.
- executor (aclOpExecutor**, output): operator executor, covering the operator computation process.

Returns:

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

The first-phase API implements input parameter verification. The following errors may be thrown:
161001 (ACLNN_ERR_PARAM_NULLPTR): 1. The input selfRef or prob is a null pointer.
161002 (ACLNN_ERR_PARAM_INVALID): 1. The data type or format of selfRef or prob is not supported.
                                      2. selfRef or prob has more than 8 dimensions.

aclnnInplaceBernoulliTensor

Parameters:
- workspace (void*, input): address of the workspace to be allocated on the device.
- workspaceSize (uint64_t, input): size of the workspace to be allocated on the device, which is obtained by calling aclnnInplaceBernoulliTensorGetWorkspaceSize.
- executor (aclOpExecutor*, input): operator executor, covering the operator computation process.
- stream (aclrtStream, input): stream for executing the task.
Returns:

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

Constraints

Deterministic computation:
- aclnnBernoulliTensor and aclnnInplaceBernoulliTensor each default to a deterministic implementation.
The values of elements in the input prob must be within the range [0, 1].
When the shapes of prob and self(or selfRef) do not match, only the corresponding elements are computed, and the behavior for the remaining data is undefined. For example, when the shape of prob is [4, 2] and that of self is [4, 4], only the first 8 elements are computed and the output shape is [4, 4]. When the shape of prob is [4, 4, 2] and that of self is [4, 4], only the first 16 elements are computed and the output shape is [4, 4].

Example

The following example is for reference only. For details, see Compilation and Running Sample.

aclnnBernoulliTensor

#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_bernoulli.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 code 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.
  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 the data from the host to the device.
  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);

  // Compute the strides of the 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 aclCreateTensor 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 code for device/stream initialization. For details, see the ACL API manual.
  // Set the device ID in use.
  int32_t deviceId = 0;
  aclrtStream stream;
  auto ret = Init(deviceId, &stream);
  // Customize the error handling as needed.
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("Init acl failed. ERROR: %d\n", ret); return ret);

  // 2. Construct the inputs and outputs based on the API definition.
  std::vector<int64_t> selfShape = {4, 4};
  std::vector<int64_t> probShape = {4, 4};
  std::vector<int64_t> outShape = {4, 4};
  void* selfDeviceAddr = nullptr;
  void* probDeviceAddr = nullptr;
  void* outDeviceAddr = nullptr;
  aclTensor* self = nullptr;
  aclTensor* prob = nullptr;
  aclTensor* out = nullptr;
  std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15};
  std::vector<float> probHostData = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
  std::vector<float> outHostData(16, 0);
  int64_t seed = 0;
  int64_t offset = 0;
  // Create a self aclTensor.
  ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create a prob aclTensor.
  ret = CreateAclTensor(probHostData, probShape, &probDeviceAddr, aclDataType::ACL_FLOAT, &prob);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create an out aclTensor.
  ret = CreateAclTensor(outHostData, outShape, &outDeviceAddr, aclDataType::ACL_FLOAT, &out);
  CHECK_RET(ret == ACL_SUCCESS, return ret);

  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;

  // aclnnBernoulliTensor API call example
  // 3. Call the CANN operator library API.
  // Call the first-phase API of aclnnBernoulliTensor.
  ret = aclnnBernoulliTensorGetWorkspaceSize(self, prob, seed, offset, out, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnBernoulliTensorGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
  // Allocate 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 aclnnBernoulliTensor.
  ret = aclnnBernoulliTensor(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnBernoulliTensor failed. ERROR: %d\n", ret); return ret);

  // 4. (Boilerplate code) Wait until the task execution is 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 to the host. Modify the code based on the 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. Destroy aclTensor and aclScalar. Modify the code based on the API definition.
  aclDestroyTensor(self);
  aclDestroyTensor(prob);
  aclDestroyTensor(out);

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

aclnnInplaceBernoulliTensor

#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_bernoulli.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 code 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.
  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 the data from the host to the device.
  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);

  // Compute the strides of the 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 aclCreateTensor 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 code for device/stream initialization. For details, see the ACL API manual.
  // Set the device ID in use.
  int32_t deviceId = 0;
  aclrtStream stream;
  auto ret = Init(deviceId, &stream);
  // Customize the error handling as needed.
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("Init acl failed. ERROR: %d\n", ret); return ret);

  // 2. Construct the inputs and outputs based on the API definition.
  std::vector<int64_t> selfShape = {4, 4};
  std::vector<int64_t> probShape = {4, 4};
  void* selfDeviceAddr = nullptr;
  void* probDeviceAddr = nullptr;
  aclTensor* self = nullptr;
  aclTensor* prob = nullptr;
  std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15};
  std::vector<float> probHostData = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
  int64_t seed = 0;
  int64_t offset = 0;
  // Create a self aclTensor.
  ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create a prob aclTensor.
  ret = CreateAclTensor(probHostData, probShape, &probDeviceAddr, aclDataType::ACL_FLOAT, &prob);
  CHECK_RET(ret == ACL_SUCCESS, return ret);

  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;

  // 3. Call the CANN operator library API.
  // Call the first-phase API of aclnnInplaceBernoulliTensor.
  ret = aclnnInplaceBernoulliTensorGetWorkspaceSize(self, prob, seed, offset, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnInplaceBernoulliTensorGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
  // Allocate 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 aclnnInplaceBernoulliTensor.
  ret = aclnnInplaceBernoulliTensor(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnInplaceBernoulliTensor failed. ERROR: %d\n", ret); return ret);

  // 4. (Boilerplate code) Wait until the task execution is 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 to the host. Modify the code based on the API definition.
  auto size = GetShapeSize(selfShape);
  std::vector<float> resultData(size, 0);
  ret = aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]), selfDeviceAddr,
                    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. Destroy aclTensor and aclScalar. Modify the code based on the API definition.
  aclDestroyTensor(self);
  aclDestroyTensor(prob);

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