下载
中文
注册

aclnnLeakyRelu&aclnnInplaceLeakyRelu

支持的产品型号

  • Atlas 训练系列产品
  • Atlas A2 训练系列产品/Atlas 800I A2 推理产品

接口原型

  • aclnnLeakyRelu和aclnnInplaceLeakyRelu实现相同的功能,使用区别如下,请根据自身实际场景选择合适的算子。

    • aclnnLeakyRelu:需新建一个输出张量对象存储计算结果。
    • aclnnInplaceLeakyRelu:无需新建输出张量对象,直接在输入张量的内存中存储计算结果。
  • 每个算子分为两段式接口,必须先调用“aclnnLeakyReluGetWorkspaceSizee”或者“aclnnInplaceLeakyReluGetWorkspaceSize”口获取计算所需workspace大小以及包含了算子计算流程的执行器,再调用“aclnnLeakyRelu”或者“aclnnInplaceLeakyRelu”接口执行计算。

    • aclnnStatus aclnnLeakyReluGetWorkspaceSize(const aclTensor *self, const aclScalar *negativeSlope, aclTensor *out, uint64_t *workspaceSize, aclOpExecutor **executor)
    • aclnnStatus aclnnLeakyRelu(void *workspace, uint64_t workspaceSize, aclOpExecutor *executor, aclrtStream stream)
    • aclnnStatus aclnnInplaceLeakyReluGetWorkspaceSize(aclTensor *selfRef, const aclScalar *negativeSlope, uint64_t *workspaceSize, aclOpExecutor **executor)
    • aclnnStatus aclnnInplaceLeakyRelu(void *workspace, uint64_t workspaceSize, aclOpExecutor *executor, aclrtStream stream)

功能描述

  • 算子功能:激活函数。
  • 计算公式:out=max(0,self)+negativeSlopemin(0,self)out = max(0,self) + negativeSlope * min(0,self)

aclnnLeakyReluGetWorkspaceSize

  • 参数说明:

    • self(aclTensor*, 计算输入):公式中的输入self,支持非连续的Tensor数据格式支持ND。
      • Atlas 训练系列产品:数据类型支持FLOAT、FLOAT16、DOUBLE。
      • Atlas A2 训练系列产品/Atlas 800I A2 推理产品:数据类型支持FLOAT、FLOAT16、BFLOAT16、DOUBLE。
    • negativeSlope(aclScalar*, 计算输入):公式中的negativeSlope,表示self<0时的斜率,数据类型支持FLOAT。
    • out(aclTensor*, 计算输出):公式中的out,数据类型与self一致,shape与self相同。支持非连续的Tensor数据格式支持ND。
      • Atlas 训练系列产品:数据类型支持FLOAT、FLOAT16、DOUBLE。
      • Atlas A2 训练系列产品/Atlas 800I A2 推理产品:数据类型支持FLOAT、FLOAT16、BFLOAT16、DOUBLE。
    • workspaceSize(uint64_t*, 出参):返回需要在Device侧申请的workspace大小。
    • executor(aclOpExecutor**, 出参):返回op执行器,包含了算子计算流程。
  • 返回值:

    aclnnStatus:返回状态码,具体参见aclnn返回码

    161001 (ACLNN_ERR_PARAM_NULLPTR): 1. 传入的self、negativeSlope或out是空指针。
    161002 (ACLNN_ERR_PARAM_INVALID): 1. self的数据类型不在支持的范围之内。
                                      2. self的shape超过8维。
                                      3. out的shape与self不一致。

aclnnLeakyRelu

  • 参数说明:

    • workspace(void*, 入参):在Device侧申请的workspace内存地址。
    • workspaceSize(uint64_t, 入参):在Device侧申请的workspace大小,由第一段接口aclnnLeakyReluGetWorkspaceSize获取。
    • executor(aclOpExecutor*, 入参):op执行器,包含了算子计算流程。
    • stream(aclrtStream, 入参):指定执行任务的AscendCL Stream流。
  • 返回值:

    aclnnStatus:返回状态码,具体参见aclnn返回码

aclnnInplaceLeakyReluGetWorkspaceSize

  • 参数说明:

    • selfRef(aclTensor*, 计算输入|计算输出):输入输出tensor,即公式中的self与out。支持非连续的Tensor数据格式支持ND。
      • Atlas 训练系列产品:数据类型支持FLOAT、FLOAT16、DOUBLE。
      • Atlas A2 训练系列产品/Atlas 800I A2 推理产品:数据类型支持FLOAT、FLOAT16、DOUBLE。
    • negativeSlope(aclScalar*, 计算输入):公式中的negativeSlope,表示self<0时的斜率,数据类型支持FLOAT。
    • workspaceSize(uint64_t*, 出参):返回需要在Device侧申请的workspace大小。
    • executor(aclOpExecutor**, 出参):返回op执行器,包含了算子计算流程。
  • 返回值:

    aclnnStatus:返回状态码,具体参见aclnn返回码

    161001 (ACLNN_ERR_PARAM_NULLPTR): 1. 传入的selfRef、negativeSlope是空指针。
    161002 (ACLNN_ERR_PARAM_INVALID): 1. selfRef的数据类型不在支持的范围之内。

aclnnInplaceLeakyRelu

  • 参数说明:

    • workspace(void*, 入参):在Device侧申请的workspace内存地址。
    • workspaceSize(uint64_t, 入参):在Device侧申请的workspace大小,由第一段接口aclnnInplaceLeakyReluGetWorkspaceSize获取。
    • executor(aclOpExecutor*, 入参):op执行器,包含了算子计算流程。
    • stream(aclrtStream, 入参):指定执行任务的AscendCL Stream流。
  • 返回值:

    aclnnStatus:返回状态码,具体参见aclnn返回码

约束与限制

调用示例

示例代码如下,仅供参考,具体编译和执行过程请参考编译与运行样例

aclnnLeakyRelu示例代码:

#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_leaky_relu.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) {
  // 固定写法,AscendCL初始化
  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);
  // 调用aclrtMalloc申请device侧内存
  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);
  // 调用aclrtMemcpy将host侧数据拷贝到device侧内存上
  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);

  // 计算连续tensor的strides
  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];
  }

  // 调用aclCreateTensor接口创建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. (固定写法)device/stream初始化,参考AscendCL对外接口列表
  // 根据自己的实际device填写deviceId
  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. 构造输入与输出,需要根据API的接口自定义构造
  std::vector<int64_t> selfShape = {4};
  std::vector<int64_t> outShape = {4};
  void* selfDeviceAddr = nullptr;
  void* outDeviceAddr = nullptr;
  aclTensor* self = nullptr;
  aclScalar* negativeSlope = nullptr;
  aclTensor* out = nullptr;
  std::vector<float> selfHostData = {1, 2, 3, 4};
  std::vector<float> outHostData = {0, 0, 0, 0};
  float negativeSlopeValue = 0.01f;
  // 创建self aclTensor
  ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // 创建negativeSlope aclScalar
  negativeSlope = aclCreateScalar(&negativeSlopeValue, aclDataType::ACL_FLOAT);
  CHECK_RET(negativeSlope != nullptr, LOG_PRINT("negativeSlope is null!\n", negativeSlope); return false);
  // 创建out aclTensor
  ret = CreateAclTensor(outHostData, outShape, &outDeviceAddr, aclDataType::ACL_FLOAT, &out);
  CHECK_RET(ret == ACL_SUCCESS, return ret);

  // 3. 调用CANN算子库API,需要修改为具体的API名称
  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;
  // 调用aclnnLeakyRelu第一段接口
  ret = aclnnLeakyReluGetWorkspaceSize(self, negativeSlope, out, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnLeakyReluGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
  // 根据第一段接口计算出的workspaceSize申请device内存
  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);
  }
  // 调用aclnnLeakyRelu第二段接口
  ret = aclnnLeakyRelu(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnLeakyRelu failed. ERROR: %d\n", ret); return ret);

  // 4. (固定写法)同步等待任务执行结束
  ret = aclrtSynchronizeStream(stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSynchronizeStream failed. ERROR: %d\n", ret); return ret);

  // 5. 获取输出的值,将device侧内存上的结果拷贝至host侧,需要根据具体API的接口定义修改
  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. 释放aclTensor和aclScalar,需要根据具体API的接口定义修改
  aclDestroyTensor(self);
  aclDestroyScalar(negativeSlope);
  aclDestroyTensor(out);
    
  // 7. 释放device资源,需要根据具体API的接口定义修改
  aclrtFree(selfDeviceAddr);
  aclrtFree(outDeviceAddr);
  if (workspaceSize > 0) {
    aclrtFree(workspaceAddr);
  }
  aclrtDestroyStream(stream);
  aclrtResetDevice(deviceId);
  aclFinalize();
  return 0;
}

aclnnInplaceLeakyRelu示例代码:

#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_leaky_relu.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) {
  // 固定写法,AscendCL初始化
  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);
  // 调用aclrtMalloc申请device侧内存
  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);
  // 调用aclrtMemcpy将host侧数据拷贝到device侧内存上
  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);

  // 计算连续tensor的strides
  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];
  }

  // 调用aclCreateTensor接口创建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. (固定写法)device/stream初始化,参考AscendCL对外接口列表
  // 根据自己的实际device填写deviceId
  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. 构造输入与输出,需要根据API的接口自定义构造
  std::vector<int64_t> selfRefShape = {4};
  void* selfRefDeviceAddr = nullptr;
  aclTensor* selfRef = nullptr;
  aclScalar* negativeSlope = nullptr;
  std::vector<float> selfRefHostData = {1, 2, 3, 4};
  float negativeSlopeValue = 0.01f;
  // 创建selfRef aclTensor
  ret = CreateAclTensor(selfRefHostData, selfRefShape, &selfRefDeviceAddr, aclDataType::ACL_FLOAT, &selfRef);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // 创建negativeSlope aclScalar
  negativeSlope = aclCreateScalar(&negativeSlopeValue, aclDataType::ACL_FLOAT);
  CHECK_RET(negativeSlope != nullptr, LOG_PRINT("negativeSlope is null!\n", negativeSlope); return false);

  // 3. 调用CANN算子库API,需要修改为具体的API名称
  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;
  // 调用aclnnInplaceLeakyRelu第一段接口
  ret = aclnnInplaceLeakyReluGetWorkspaceSize(selfRef, negativeSlope, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnInplaceLeakyReluGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
  // 根据第一段接口计算出的workspaceSize申请device内存
  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);
  }
  // 调用aclnnInplaceLeakyRelu第二段接口
  ret = aclnnInplaceLeakyRelu(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnInplaceLeakyRelu failed. ERROR: %d\n", ret); return ret);

  // 4. (固定写法)同步等待任务执行结束
  ret = aclrtSynchronizeStream(stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSynchronizeStream failed. ERROR: %d\n", ret); return ret);

  // 5. 获取输出的值,将device侧内存上的结果拷贝至host侧,需要根据具体API的接口定义修改
  auto size = GetShapeSize(selfRefShape);
  std::vector<float> resultData(size, 0);
  ret = aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]), selfRefDeviceAddr,
                    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. 释放aclTensor和aclScalar,需要根据具体API的接口定义修改
  aclDestroyTensor(selfRef);
  aclDestroyScalar(negativeSlope);

  // 7. 释放device资源,需要根据具体API的接口定义修改
  aclrtFree(selfRefDeviceAddr);
  if (workspaceSize > 0) {
    aclrtFree(workspaceAddr);
  }
  aclrtDestroyStream(stream);
  aclrtResetDevice(deviceId);
  aclFinalize();
  return 0;
}