下载
中文
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

aclnnClampMinTensor&aclnnInplaceClampMinTensor

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

  • aclnnClampMinTensor和aclnnInplaceClampMinTensor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。

    • aclnnClampMinTensor:需新建一个输出张量对象存储计算结果。
    • aclnnInplaceClampMinTensor: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“aclnnClampMinTensorGetWorkspaceSize”或者“aclnnInplaceClampMinTensorGetWorkspaceSize”接口获取计算所需workspace大小以及包含了算子计算流程的执行器,再调用“aclnnClampMinTensor”或者“aclnnInplaceClampMinTensor”接口执行计算。

    • aclnnStatus aclnnClampMinTensorGetWorkspaceSize(const aclTensor* self, const aclTensor* clipValueMin, aclTensor* out, uint64_t* workspaceSize, aclOpExecutor** executor)
    • aclnnStatus aclnnClampMinTensor(void *workspace, uint64_t workspaceSize, aclOpExecutor *executor, aclrtStream stream)
    • aclnnStatus aclnnInplaceClampMinTensorGetWorkspaceSize(aclTensor* selfRef, const aclTensor* clipValueMin, uint64_t *workspaceSize, aclOpExecutor **executor)
    • aclnnStatus aclnnInplaceClampMinTensor(void *workspace, uint64_t workspaceSize, aclOpExecutor *executor, aclrtStream stream)

功能描述

  • 算子功能:将输入的所有元素限制在[min, inf]范围内。

  • 计算公式:

    yi=max(xi,min_valuei){y}_{i} = max({{x}_{i}},{min\_value}_{i})

aclnnClampMinTensorGetWorkspaceSize

  • 参数说明

    • self(const aclTensor *): 输入tensor,数据类型支持FLOAT16、FLOAT、FLOAT64、INT8、UINT8、INT16、INT32、INT64、BFLOAT16(Atlas A2训练系列产品/Atlas 800I A2推理产品)。支持非连续的Tensor数据格式支持ND。
    • clipValueMin(const aclTensor*): 输入下限值tensor,数据类型支持FLOAT16、FLOAT、FLOAT64、INT8、UINT8、INT16、INT32、INT64、BFLOAT16(Atlas A2训练系列产品/Atlas 800I A2推理产品),且数据类型需要与self满足数据类型推导规则(参见互推导关系)。shape需要与self满足broadcast关系,支持非连续的Tensor数据格式支持ND。
    • out(aclTensor *):输出,数据类型支持FLOAT16、FLOAT、FLOAT64、INT8、UINT8、INT16、INT32、INT64、BFLOAT16(Atlas A2训练系列产品/Atlas 800I A2推理产品),且数据类型和self保持一致,shape需要与self满足broadcast关系数据格式支持ND。
    • workspaceSize(uint64_t *):返回需要在Device侧申请的workspace大小。
    • executor(aclOpExecutor **):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

    返回161001(ACLNN_ERR_PARAM_NULLPTR):1. 传入的self、out其中一个为空指针,min为空指针。
    返回161002(ACLNN_ERR_PARAM_INVALID):1. self、out的数据类型和数据格式不在支持的范围之内。
                                          2. self的数据类型与输出out的类型不一致。

aclnnClampMinTensor

  • 参数说明:

    • workspace(void *):在Device侧申请的workspace内存地址。
    • workspaceSize(uint64_t):在Device侧申请的workspace大小,由第一段接口aclnnClampMinTensorGetWorkspaceSize获取。
    • executor(aclOpExecutor*):op执行器,包含了算子计算流程。
    • stream(aclrtStream):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

aclnnInplaceClampMinTensorGetWorkspaceSize

  • 参数说明:

    • selfRef(const aclTensor *): 输入输出tensor,数据类型支持FLOAT16、FLOAT、FLOAT64、INT8、UINT8、INT16、INT32、INT64、BFLOAT16(Atlas A2训练系列产品/Atlas 800I A2推理产品),且数据类型与clipValueMin的数据类型需满足数据类型推导规则(参见互推导关系),且推导后的数据类型需要能转换成selfRef自身的数据类型。支持非连续的Tensor数据格式支持ND。
    • clipValueMin(const aclTensor*): 输入下限值tensor,数据类型支持FLOAT16、FLOAT、FLOAT64、INT8、UINT8、INT16、INT32、INT64、BFLOAT16(Atlas A2训练系列产品/Atlas 800I A2推理产品),且数据类型需要与selfRef满足数据类型推导规则(参见互推导关系)。shape需要与self满足broadcast关系,支持非连续的Tensor数据格式支持ND,维度不超过8维。
    • workspaceSize(uint64_t *):返回需要在Device侧申请的workspace大小。
    • executor(aclOpExecutor **):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

    返回161001(ACLNN_ERR_PARAM_NULLPTR):1. 传入的selfRef为空指针,min为空指针。
    返回161002(ACLNN_ERR_PARAM_INVALID):1. selfRef的数据类型和数据格式不在支持的范围之内。
                                          2. selfRef和min的shape不满足broadcast关系。
                                          3. selfRef和min类型推导失败。
                                          4. selfRef和min的维度大小超过8。

aclnnInplaceClampMinTensor

  • 参数说明:

    • workspace(void *):在Device侧申请的workspace内存地址。
    • workspaceSize(uint64_t):在Device侧申请的workspace大小,由第一段接口aclnnInplaceClampMinTensorGetWorkspaceSize获取。
    • executor(aclOpExecutor *):op执行器,包含了算子计算流程。
    • stream(aclrtStream):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

aclnnClampMinTensor示例代码:

#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_clamp.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 shape_size = 1;
  for (auto i : shape) {
    shape_size *= i;
  }
  return shape_size;
}

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 == 0, 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, 2};
  std::vector<int64_t> minShape = {4, 2};
  std::vector<int64_t> outShape = {4, 2};
  void* selfDeviceAddr = nullptr;
  void* minDeviceAddr = nullptr;
  void* outDeviceAddr = nullptr;
  aclTensor* self = nullptr;
  aclTensor* min = nullptr;
  aclTensor* out = nullptr;
  std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7};
  std::vector<float> minHostData = {2, 1, 1, 2, 2, 6, 6, 9};
  std::vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0};
  // 创建self aclTensor
  ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // 创建min aclTensor
  ret = CreateAclTensor(minHostData, minShape, &minDeviceAddr, aclDataType::ACL_FLOAT, &min);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // 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
  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;
  // 调用aclnnClampMinTensor第一段接口
  ret = aclnnClampMinTensorGetWorkspaceSize(self, min, out, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnClampMinTensorGetWorkspaceSize 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aclnnClampMinTensor第二段接口
  ret = aclnnClampMinTensor(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnClampMinTensor 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(float),
                    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);
  aclDestroyTensor(min);
  aclDestroyTensor(out);

  // 7. 释放device资源
  aclrtFree(selfDeviceAddr);
  aclrtFree(minDeviceAddr);
  aclrtFree(outDeviceAddr);
  if (workspaceSize > 0) {
    aclrtFree(workspaceAddr);
  }
  aclrtDestroyStream(stream);
  aclrtResetDevice(deviceId);
  aclFinalize();
  return 0;
}

aclnnInplaceClampMinTensor示例代码:

#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_clamp.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
  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, 2};
  std::vector<int64_t> minShape = {4, 2};
  void* selfDeviceAddr = nullptr;
  void* minDeviceAddr = nullptr;
  aclTensor* self = nullptr;
  aclTensor* min = nullptr;
  std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7};
  std::vector<float> minHostData = {2, 1, 1, 2, 2, 6, 6, 9};

  // 创建self aclTensor
  ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // 创建min aclTensor
  ret = CreateAclTensor(minHostData, minShape, &minDeviceAddr, aclDataType::ACL_FLOAT, &min);
  CHECK_RET(ret == ACL_SUCCESS, return ret);

  // 3. 调用CANN算子库API
  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;
  // 调用aclnnInplaceClampMinTensor第一段接口
  ret = aclnnInplaceClampMinTensorGetWorkspaceSize(self, min, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnInplaceClampMinTensorGetWorkspaceSize 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aclnnInplaceClampMinTensor第二段接口
  ret = aclnnInplaceClampMinTensor(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnInplaceClampMinTensor 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
  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. 释放aclTensor和aclScalar
  aclDestroyTensor(self);
  aclDestroyTensor(min);

  // 7. 释放device资源
  aclrtFree(selfDeviceAddr);
  aclrtFree(minDeviceAddr);
  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