aclnnTransConvolutionWeight
支持的产品型号
- Atlas 推理系列产品。
接口原型
每个算子分为两段式接口，必须先调用“aclnnTransConvolutionWeightGetWorkspaceSize”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“aclnnTransConvolutionWeight”接口执行计算。
aclnnStatus aclnnTransConvolutionWeightGetWorkspaceSize(const aclTensor* weightIn, bool transposed, const int64_t groups, aclTensor* weightOut, uint64_t* workspaceSize, aclOpExecutor** executor)aclnnStatus aclnnTransConvolutionWeight(void* workspace, uint64_t workspaceSize, aclOpExecutor* executor, aclrtStream stream)
功能描述
需要和aclnnCalculateConvolutionWeightSize接口配套使用，用于创建一个对于Convolution算子计算性能亲和的weight Tensor。
aclnnTransConvolutionWeightGetWorkspaceSize
参数说明：
- weightIn(aclTensor *, 计算输入)：Device侧的aclTensor，表示一个待处理的Convolution的weightTensor，数据类型支持FLOAT16、FLOAT32（仅Atlas 推理系列产品支持），支持非连续的Tensor，数据格式支持NCHW。
- transposed(bool *, 计算输入)：Host侧的布尔值，表明是否为转置卷积，目前仅支持设为false。
- groups(int64_t *, 计算输入)：Host侧的整型，数据类型支持INT64，表示从输入通道到输出通道的块链接个数。
- weightOut(aclTensor *, 计算输出)：Device侧的aclTensor，表示返回输入weight转换为私有格式后的tensor，数据类型支持FLOAT16，不支持非连续的Tensor，数据格式支持NCHW。
- 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. 输入是空指针。 161002(ACLNN_ERR_PARAM_INVALID)：1. 输入输出Tensor的数据类型、数据格式以及其他参数不符合预期。比如输入weightIn为非FLOAT16、FLOAT32数据类型或者非NCHW数据格式。
aclnnTransConvolutionWeight
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aclnnTransConvolutionWeightGetWorkspaceSize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
- 仅支持正向Conv2D场景。
- 不支持转置卷积。
- 不支持cache缓存能力。
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 <memory>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_convolution.h"
#include "aclnnop/aclnn_trans_convolution_weight.h"
using namespace std;
#define CHECK_RET(cond, return_expr) \
do { \
if (!(cond)) { \
return_expr; \
} \
} while (0)
#define CHECK_FREE_RET(cond, return_expr) \
do { \
if (!(cond)) { \
Finalize(deviceId, stream); \
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_NCHW,
shape.data(), shape.size(), *deviceAddr);
return 0;
}
template <typename T>
int CreateWeightAclTensor(const std::vector<T> &hostData, const std::vector<int64_t> &shape, void **deviceAddr,
aclDataType dataType, aclTensor **tensor, uint64_t &TransWeightSize)
{
auto size = GetShapeSize(shape) * sizeof(T);
// 调用transweight host接口 计算实际elements数量
aclIntArray* weightSize = aclCreateIntArray(shape.data(), shape.size());
std::unique_ptr<aclIntArray, aclnnStatus (*)(const aclIntArray *)> weightSizePtr(weightSize, aclDestroyIntArray);
auto ret = aclnnCalculateConvolutionWeightSize(weightSize, false, 1, aclDataType::ACL_FLOAT16, &TransWeightSize);
// 调用aclrtMalloc申请device侧内存
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_NCHW,
shape.data(), shape.size(), *deviceAddr);
return 0;
}
void Finalize(int32_t deviceId, aclrtStream& stream)
{
aclrtDestroyStream(stream);
aclrtResetDevice(deviceId);
aclFinalize();
}
int aclnnTransConvolutionWeightTest(int32_t deviceId, aclrtStream& stream) {
auto ret = Init(deviceId, &stream);
CHECK_FREE_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> inputShape = {1, 4, 16, 16};
std::vector<int64_t> weightShape = {2, 4, 8, 8};
std::vector<int64_t> biasShape = {2};
std::vector<int64_t> outShape = {1, 2, 9, 9};
void* inputDeviceAddr = nullptr;
void* weightDeviceAddr = nullptr;
void* biasDeviceAddr = nullptr;
void* outDeviceAddr = nullptr;
aclTensor* input = nullptr;
aclTensor* weight = nullptr;
aclTensor* bias = nullptr;
aclTensor* out = nullptr;
std::vector<float> inputHostData(1024, 1);
std::vector<float> weightHostData(512, 1);
std::vector<float> biasHostData(2, 1);
std::vector<float> outHostData(162, 0);
uint64_t transWeightSize = 0;
// 创建input aclTensor
ret = CreateAclTensor(inputHostData, inputShape, &inputDeviceAddr, aclDataType::ACL_FLOAT, &input);
std::unique_ptr<aclTensor, aclnnStatus (*)(const aclTensor *)> selfTensorPtr(input, aclDestroyTensor);
std::unique_ptr<void, aclError (*)(void *)> inputDeviceAddrPtr(inputDeviceAddr, aclrtFree);
CHECK_FREE_RET(ret == ACL_SUCCESS, return ret);
// 创建weight aclTensor
ret = CreateWeightAclTensor(weightHostData, weightShape, &weightDeviceAddr, aclDataType::ACL_FLOAT,
&weight, transWeightSize);
std::unique_ptr<aclTensor, aclnnStatus (*)(const aclTensor *)> weightTensorPtr(weight, aclDestroyTensor);
std::unique_ptr<void, aclError (*)(void *)> weightDeviceAddrPtr(weightDeviceAddr, aclrtFree);
CHECK_FREE_RET(ret == ACL_SUCCESS, return ret);
// 创建bias aclTensor
ret = CreateAclTensor(biasHostData, biasShape, &biasDeviceAddr, aclDataType::ACL_FLOAT, &bias);
std::unique_ptr<aclTensor, aclnnStatus (*)(const aclTensor *)> biasTensorPtr(bias, aclDestroyTensor);
std::unique_ptr<void, aclError (*)(void *)> biasDeviceAddrPtr(biasDeviceAddr, aclrtFree);
CHECK_FREE_RET(ret == ACL_SUCCESS, return ret);
// 创建out aclTensor
ret = CreateAclTensor(outHostData, outShape, &outDeviceAddr, aclDataType::ACL_FLOAT, &out);
std::unique_ptr<aclTensor, aclnnStatus (*)(const aclTensor *)> outTensorPtr(out, aclDestroyTensor);
std::unique_ptr<void, aclError (*)(void *)> outDeviceAddrPtr(outDeviceAddr, aclrtFree);
CHECK_FREE_RET(ret == ACL_SUCCESS, return ret);
// 创建Transweight acltensor
void* transWeightDeviceAddr = nullptr;
uint64_t size = transWeightSize * sizeof(float) / 2;
ret = aclrtMalloc(&transWeightDeviceAddr, size, ACL_MEM_MALLOC_HUGE_FIRST);
CHECK_FREE_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtMalloc failed. ERROR: %d\n", ret);return ret);
std::vector<float> transData;
transData.resize(transWeightSize * 2);
// 调用aclrtMemcpy将Host侧数据拷贝到device侧内存上transData.data()
ret = aclrtMemcpy(transWeightDeviceAddr, size, transData.data(), size, ACL_MEMCPY_HOST_TO_DEVICE);
CHECK_FREE_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtMemcpy failed. ERROR: %d\n", ret);
return ret);
// 计算连续tensor的strides
vector<int64_t> shape = weightShape;
std::vector<int64_t> s(shape.size(), 1);
for (int64_t i = shape.size() - 2; i >= 0; i--) {
s[i] = shape[i + 1] * s[i + 1];
}
aclTensor* transWeight = aclCreateTensor(shape.data(), shape.size(), aclDataType::ACL_FLOAT16, s.data(), 0, aclFormat::ACL_FORMAT_NCHW,
shape.data(), shape.size(), transWeightDeviceAddr);
std::unique_ptr<aclTensor, aclnnStatus (*)(const aclTensor *)> transWeightTensorPtr(transWeight, aclDestroyTensor);
std::unique_ptr<void, aclError (*)(void *)> transWeightDeviceAddrAddrPtr(transWeightDeviceAddr, aclrtFree);
// 3. 调用CANN算子库API，需要修改为具体的Api名称
int8_t cubeMathType = 2; // USE_FP16
uint64_t workspaceSize = 0;
aclOpExecutor* executor;
bool transposed = 0;
uint64_t groups = 1;
// 调用TransWeight
ret = aclnnTransConvolutionWeightGetWorkspaceSize(weight, transposed, groups, transWeight,
&workspaceSize, &executor);
CHECK_FREE_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnTransConvolutionWeightGetWorkspaceSize 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;
std::unique_ptr<void, aclError (*)(void *)> workspaceAddrPtr(nullptr, aclrtFree);
if (workspaceSize > 0) {
ret = aclrtMalloc(&workspaceAddr, workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST);
CHECK_FREE_RET(ret == ACL_SUCCESS, LOG_PRINT("allocate workspace failed. ERROR: %d\n", ret); return ret);
workspaceAddrPtr.reset(workspaceAddr);
}
// 调用aclnnTransConvolutionWeight第二段接口
ret = aclnnTransConvolutionWeight(workspaceAddr, workspaceSize, executor, stream);
CHECK_FREE_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnTransConvolutionWeight failed. ERROR: %d\n", ret); return ret);
std::vector<int64_t> convStrides = {1, 1, 1, 1};
std::vector<int64_t> convPads = {0, 0, 0, 0};
std::vector<int64_t> convOutPads = {1, 1, 1, 1};
std::vector<int64_t> convDilations = {1, 1, 1, 1};
aclIntArray *strides = aclCreateIntArray(convStrides.data(), 2);
std::unique_ptr<aclIntArray, aclnnStatus (*)(const aclIntArray *)> stridesPtr(strides, aclDestroyIntArray);
CHECK_FREE_RET(strides != nullptr, return ACL_ERROR_INTERNAL_ERROR);
aclIntArray *pads = aclCreateIntArray(convPads.data(), 2);
std::unique_ptr<aclIntArray, aclnnStatus (*)(const aclIntArray *)> padsPtr(pads, aclDestroyIntArray);
CHECK_FREE_RET(pads != nullptr, return ACL_ERROR_INTERNAL_ERROR);
aclIntArray *outPads = aclCreateIntArray(convOutPads.data(), 2);
std::unique_ptr<aclIntArray, aclnnStatus (*)(const aclIntArray *)> outPadsPtr(outPads, aclDestroyIntArray);
CHECK_FREE_RET(outPads != nullptr, return ACL_ERROR_INTERNAL_ERROR);
aclIntArray *dilations = aclCreateIntArray(convDilations.data(), 2);
std::unique_ptr<aclIntArray, aclnnStatus (*)(const aclIntArray *)> dilationsPtr(dilations, aclDestroyIntArray);
CHECK_FREE_RET(dilations != nullptr, return ACL_ERROR_INTERNAL_ERROR);
// 3. 调用CANN算子库API，需要修改为具体的API
workspaceSize = 0;
// 调用aclnnConvolution第一段接口
ret = aclnnConvolutionGetWorkspaceSize(input, transWeight, bias, strides, pads, dilations, false, outPads, groups,
out, cubeMathType, &workspaceSize, &executor);
CHECK_FREE_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnConvolutionGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
// 根据第一段接口计算出的workspaceSize申请device内存
workspaceAddr = nullptr;
if (workspaceSize > 0) {
ret = aclrtMalloc(&workspaceAddr, workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST);
CHECK_FREE_RET(ret == ACL_SUCCESS, LOG_PRINT("allocate workspace failed. ERROR: %d\n", ret); return ret);
}
// 调用aclnnConvolution第二段接口
ret = aclnnConvolution(workspaceAddr, workspaceSize, executor, stream);
CHECK_FREE_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnConvolution 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_FREE_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
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_FREE_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]);
}
return ACL_SUCCESS;
}
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 = aclnnTransConvolutionWeightTest(deviceId, stream);
CHECK_FREE_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnTransConvolutionWeightTest failed. ERROR: %d\n", ret); return ret);
Finalize(deviceId, stream);
return 0;
}