下载
EN
注册
我要评分
文档获取效率
文档正确性
内容完整性
文档易理解
在线提单
论坛求助
昇腾小AI

aclnnBatchNormBackward

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

每个算子分为两段式接口,必须先调用“aclnnBatchNormBackwardGetWorkspaceSize”接口获取入参并根据流程计算所需workspace大小,再调用“aclnnBatchNormBackward”接口执行计算。

  • aclnnStatus aclnnBatchNormBackwardGetWorkspaceSize(const aclTensor *gradOut, const aclTensor *input, const aclTensor *weight, const aclTensor *runningMean, const aclTensor *runningVar, const aclTensor *saveMean, const aclTensor *saveInvstd, bool training, double eps, const aclBoolArray *outputMask, aclTensor *gradInput, aclTensor *gradWeight, aclTensor *gradBias, uint64_t *workspaceSize, aclOpExecutor **executor)
  • aclnnStatus aclnnBatchNormBackward(void *workspace, uint64_t workspaceSize, aclOpExecutor *executor, aclrtStream stream)

功能描述

  • 算子功能: 正则化反向计算。
  • 计算公式:
lx^i=lyiγ\frac{\partial l}{\partial \hat{x}_i} = \frac{\partial l}{\partial y_i} \cdot γ lσB2=i=0mlx^i(xiμB)12(σB2+ε)3/2\frac{\partial l}{\partial σ^2_B} = \sum^m_{i=0}\frac{\partial l}{\partial \hat{x}_i} \cdot (x_i-μ_B) \cdot \frac{-1}{2}(σ^2_B + ε)^{-3/2} lμB=(i=0mlx^i1σB2+ε)+lσB2i=0m2(xiμB)m\frac{\partial l}{\partial μ_B} = (\sum^m_{i=0}\frac{\partial l}{\partial \hat{x}_i} \cdot \frac{-1}{\sqrt{σ^2_B + ε}}) + \frac{\partial l}{\partial σ^2_B} \cdot \frac{\sum^m_{i=0}-2(x_i-μ_B)}{m} lxi=lx^i1σB2+ε)+lσB22(xiμB)m+lμB1m\frac{\partial l}{\partial x_i} = \frac{\partial l}{\partial \hat{x}_i} \cdot \frac{1}{\sqrt{σ^2_B + ε}}) + \frac{\partial l}{\partial σ^2_B} \cdot \frac{2(x_i-μ_B)}{m} + \frac{\partial l}{\partial μ_B} \cdot \frac{1}{m} lγ=i=0mlyix^\frac{\partial l}{\partial γ} = \sum^m_{i=0} \frac{\partial l}{\partial y_i} \cdot \hat{x} lβ=i=0mlyi\frac{\partial l}{\partial β} = \sum^m_{i=0} \frac{\partial l}{\partial y_i}

aclnnBatchNormBackwardGetWorkspaceSize

  • 参数说明:

    • gradOut(const aclTensor *, 计算输入): 梯度Tensor,Device侧的aclTensor,数据类型支持FLOAT,FLOAT16,BFLOAT16(仅Atlas A2训练系列产品/Atlas 800I A2推理产品支持),支持非连续的Tensor,支持的shape和格式有:2维(对应的格式为NC),3维(对应的格式为NCL),4维(对应的格式为NCHW),5维(对应的格式为NCDHW),6-8维(对应的格式为ND,其中第2维固定为channel轴)。
    • input(const aclTensor *, 计算输入): 正向的输入Tensor,Device侧的aclTensor,数据类型支持FLOAT,FLOAT16,BFLOAT16(仅Atlas A2训练系列产品/Atlas 800I A2推理产品支持),支持非连续的Tensor,支持的shape和格式有:2维(对应的格式为NC),3维(对应的格式为NCL),4维(对应的格式为NCHW),5维(对应的格式为NCDHW),6-8维(对应的格式为ND,其中第2维固定为channel轴)。
    • weight(const aclTensor *, 计算输入): 权重Tensor,Device侧的aclTensor,数据类型仅支持FLOAT,支持非连续的Tensor数据格式为ND。shape为1维,长度与input入参中channel轴的长度相等。
    • runningMean(const aclTensor *, 计算输入): 训练期间计算的平均值,Device侧的aclTensor数据类型仅支持FLOAT,支持非连续的Tensor数据格式为ND。shape为1维,长度与input入参中channel轴的长度相等。
    • runningVar(const aclTensor *, 计算输入): 训练期间计算的方差,Device侧的aclTensor,数据类型仅支持FLOAT,支持非连续的Tensor数据格式为ND。shape为1维,长度与input入参中channel轴的长度相等。
    • saveMean(const aclTensor *, 计算输入): 保存的均值,Device侧的aclTensor,数据类型仅支持FLOAT,支持非连续的Tensor数据格式为ND。shape为1维,长度与input入参中channel轴的长度相等。
    • saveInvstd(const aclTensor *, 计算输入): 保存的标准差的倒数,Device侧的aclTensor,数据类型仅支持FLOAT,支持非连续的Tensor数据格式为ND。shape为1维,长度与input入参中channel轴的长度相等。
    • training(bool, 计算输入): Host侧的bool值,标记是否训练场景,true表示训练场景,false表示推理场景。
    • eps(double *, 计算输入): Host侧的double值。添加到方差中的值,以避免出现除以零的情况。
    • outputMask(const aclBoolArray *, 计算输入): aclBoolArray类型,输出的掩码。
    • gradInput(aclTensor *, 出参): 可选输出,若outputMask[0]为True,则需要输出,否则不输出;输入Tensor的梯度,Device侧的aclTensor,数据类型支持FLOAT,FLOAT16,BFLOAT16(仅Atlas A2训练系列产品/Atlas 800I A2推理产品支持),支持非连续的Tensor,支持的shape和格式有:2维(对应的格式为NC),3维(对应的格式为NCL),4维(对应的格式为NCHW),5维(对应的格式为NCDHW),6-8维(对应的格式为ND,其中第2维固定为channel轴)。
    • gradWeight(aclTensor *, 出参): 可选输出,若outputMask[1]为True,则需要输出,否则不输出;缩放参数的梯度,Device侧的aclTensor,数据类型仅支持FLOAT,支持非连续的Tensor数据格式为ND。shape为1维,长度与input入参中channel轴的长度相等。
    • gradBias(aclTensor *, 出参): 可选输出,若outputMask[2]为True,则需要输出,否则不输出;偏置参数的梯度,Device侧的aclTensor,数据类型仅支持FLOAT,支持非连续的Tensor数据格式为ND。shape为1维,长度与input入参中channel轴的长度相等。
    • 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

第一段接口完成入参校验,出现以下场景时报错:
ACLNN_ERR_PARAM_NULLPTR:1. 传入的指针类型入参是空指针。
ACLNN_ERR_PARAM_INVALID:1. input,gradOut,gradInput数据类型和数据格式不在支持的范围之内。
                         2. input,gradOut,gradInput数据的shape不在支持的范围内。

aclnnBatchNormBackward

  • 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aclnnBatchNormBackwardGetWorkspaceSize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

#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_batch_norm_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) {
  // 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根据自己的需要处理
  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> gradOutShape = {1, 2, 4};
  std::vector<int64_t> selfShape = {1, 2, 4};
  std::vector<int64_t> weightShape = {2};
  std::vector<int64_t> rMeanShape = {2};
  std::vector<int64_t> rVarShape = {2};
  std::vector<int64_t> sMeanShape = {2};
  std::vector<int64_t> sVarShape = {2};
  std::vector<int64_t> gradInShape = {1, 2, 4};
  std::vector<int64_t> gradWeightShape = {2};
  std::vector<int64_t> gradBiasShape = {2};
  void* gradOutDeviceAddr = nullptr;
  void* selfDeviceAddr = nullptr;
  void* weightDeviceAddr = nullptr;
  void* rMeanDeviceAddr = nullptr;
  void* rVarDeviceAddr = nullptr;
  void* sMeanDeviceAddr = nullptr;
  void* sVarDeviceAddr = nullptr;
  void* outMaskDeviceAddr = nullptr;
  void* gradInDeviceAddr = nullptr;
  void* gradWeightDeviceAddr = nullptr;
  void* gradBiasDeviceAddr = nullptr;
  aclTensor* gradOut = nullptr;
  aclTensor* self = nullptr;
  aclTensor* weight = nullptr;
  aclTensor* rMean = nullptr;
  aclTensor* rVar = nullptr;
  aclTensor* sMean = nullptr;
  aclTensor* sVar = nullptr;
  aclBoolArray* outMask = nullptr;
  aclTensor* gradIn = nullptr;
  aclTensor* gradWeight = nullptr;
  aclTensor* gradBias = nullptr;
  std::vector<float> gradOutHostData = {0, 1, 2, 3, 4, 5, 6, 7};
  std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7};
  std::vector<float> weightHostData = {1, 1};
  std::vector<float> rMeanHostData = {0, 0};
  std::vector<float> rVarHostData = {1, 1};
  std::vector<float> sMeanHostData = {0, 0};
  std::vector<float> sVarHostData = {1, 1};
  std::vector<float> gradInHostData(8, 0);
  std::vector<float> gradWeightHostData(2, 0);
  std::vector<float> gradBiasHostData(2, 0);;
  bool training = true;
  double eps = 1e-5;
  // 创建gradOut aclTensor
  ret = CreateAclTensor(gradOutHostData, gradOutShape, &gradOutDeviceAddr, aclDataType::ACL_FLOAT, &gradOut);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // 创建self aclTensor
  ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // 创建weight aclTensor
  ret = CreateAclTensor(weightHostData, weightShape, &weightDeviceAddr, aclDataType::ACL_FLOAT, &weight);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // 创建rMean aclTensor
  ret = CreateAclTensor(rMeanHostData, rMeanShape, &rMeanDeviceAddr, aclDataType::ACL_FLOAT, &rMean);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // 创建rVar aclTensor
  ret = CreateAclTensor(rVarHostData, rVarShape, &rVarDeviceAddr, aclDataType::ACL_FLOAT, &rVar);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // 创建sMean aclTensor
  ret = CreateAclTensor(sMeanHostData, sMeanShape, &sMeanDeviceAddr, aclDataType::ACL_FLOAT, &sMean);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // 创建sVar aclTensor
  ret = CreateAclTensor(sVarHostData, sVarShape, &sVarDeviceAddr, aclDataType::ACL_FLOAT, &sVar);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // 创建outMask aclBoolArray
  bool maskData[2] = {true, true};
  outMask = aclCreateBoolArray(&(maskData[0]), 2);
  // 创建gradIn aclTensor
  ret = CreateAclTensor(gradInHostData, gradInShape, &gradInDeviceAddr, aclDataType::ACL_FLOAT, &gradIn);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // 创建gradWeight aclTensor
  ret = CreateAclTensor(gradWeightHostData, gradWeightShape, &gradWeightDeviceAddr, aclDataType::ACL_FLOAT, &gradWeight);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // 创建gradBias aclTensor
  ret = CreateAclTensor(gradBiasHostData, gradBiasShape, &gradBiasDeviceAddr, aclDataType::ACL_FLOAT, &gradBias);
  CHECK_RET(ret == ACL_SUCCESS, return ret);

  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;

  // aclnnBatchNormBackward接口调用示例
  // 3. 调用CANN算子库API,需要修改为具体的API名称
  // 调用aclnnBatchNormBackward第一段接口
  ret = aclnnBatchNormBackwardGetWorkspaceSize(gradOut, self, weight, rMean, rVar, sMean, sVar, training, eps, outMask, gradIn, gradWeight, gradBias, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnBatchNormBackwardGetWorkspaceSize 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aclnnBatchNormBackward第二段接口
  ret = aclnnBatchNormBackward(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnBatchNormBackward 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(gradInShape);
  std::vector<float> resultData(size, 0);
  ret = aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]), gradInDeviceAddr,
                    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(gradOut);
  aclDestroyTensor(self);
  aclDestroyTensor(weight);
  aclDestroyTensor(rMean);
  aclDestroyTensor(rVar);
  aclDestroyTensor(sMean);
  aclDestroyTensor(sVar);
  aclDestroyBoolArray(outMask);
  aclDestroyTensor(gradIn);
  aclDestroyTensor(gradWeight);
  aclDestroyTensor(gradBias);

  // 7. 释放device资源,需要根据具体API的接口定义修改
  aclrtFree(gradOutDeviceAddr);
  aclrtFree(selfDeviceAddr);
  aclrtFree(weightDeviceAddr);
  aclrtFree(rMeanDeviceAddr);
  aclrtFree(rVarDeviceAddr);
  aclrtFree(sMeanDeviceAddr);
  aclrtFree(sVarDeviceAddr);
  aclrtFree(outMaskDeviceAddr);
  aclrtFree(gradInDeviceAddr);
  aclrtFree(gradWeightDeviceAddr);
  aclrtFree(gradBiasDeviceAddr);
  if (workspaceSize > 0) {
    aclrtFree(workspaceAddr);
  }
  aclrtDestroyStream(stream);
  aclrtResetDevice(deviceId);
  aclFinalize();
  return 0;
}
搜索结果
找到“0”个结果

当前产品无相关内容

未找到相关内容,请尝试其他搜索词