aclnnRoiAlign
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。
- 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“aclnnRoiAlignGetWorkspaceSize”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“aclnnRoiAlign”接口执行计算。
aclnnStatus aclnnRoiAlignGetWorkspaceSize(const aclTensor* self, const aclTensor* rois, const aclTensor* batchIndices, const char* mode, int outputHeight, int outputWidth, int samplingRatio, float spatialScale, aclTensor* out, uint64_t* workspaceSize, aclOpExecutor** executor)
aclnnStatus aclnnRoiAlign(void *workspace, uint64_t workspaceSize, aclOpExecutor *executor, const aclrtStream stream)
功能描述
算子功能:RoiAlign是一种池化层,用于非均匀输入尺寸的特征图,并输出固定尺寸的特征图。
aclnnRoiAlignGetWorkspaceSize
参数说明:
self(aclTensor*, 计算输入):Device侧的aclTensor,数据类型支持FLOAT、FLOAT16。支持非连续的Tensor,数据格式支持NCHW。维度为4维,shape为(N, C, H, W)。
rois(aclTensor*, 计算输入):感兴趣区域。Device侧的aclTensor,数据类型支持FLOAT、FLOAT16。支持非连续的Tensor,数据格式支持ND。维度为2维,shape为(numRois, 4)。
batchIndices(aclTensor*, 计算输入):Device侧的aclTensor,数据类型支持INT32。支持非连续的Tensor,数据格式支持ND。表示每batch对应图像的索引。维度为1维,shape为(numRois,)。
out(aclTensor*, 计算输出):Device侧的aclTensor,数据类型支持FLOAT、FLOAT16。支持非连续的Tensor,数据格式支持NCHW。维度为4维,shape为(numRois, C, outputHeight, outputWidth)。
mode(string, 计算输入):支持"avg"和"max"。池化模式。
outputHeight(int, 计算输入):默认值为1。输出图像的高度。
outputWidth(int, 计算输入):默认值为1。输出图像的宽度。
samplingRatio(int, 计算输入):默认值为0。用于计算每个输出元素的和W上的bin数。
spatialScale(float, 计算输入):默认值为1.0。乘法空间尺度因子,将ROI坐标从其输入空间尺度转换为池化时使用的尺度,即输入特征图X相对于输入图像的空间尺度。
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。
返回码:
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、rois、batchIndices、out是空指针。
返回 161002(ACLNN_ERR_PARAM_INVALID):1. self和out的数据类型不在支持的范围内:
1) self、rois和out仅支持FLOAT、FLOAT16;
2) batchIndices仅支持INT32;
3) self、rois和out的数据类型不一致。
2. self、rois、batchIndices和out的数据格式不在支持的范围内:
1) self和out支持NCHW;
2) rois和batchIndices支持ND。
3. self、rois、batchIndices和out的shape不在支持的范围内。
1) self和out需为4维;
2) rois需为2维;
3) batchIndices需为1维。
4. mode仅支持"avg"和"max"两种取值。
5. samplingRatio需大于等于0。
6. spatialScale需大于0。
aclnnRoiAlign
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aclnnRoiAlignGetWorkspaceSize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。
返回码:
aclnnStatus:返回状态码,具体参见aclnn返回码。
约束与限制
无。
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_roi_align.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;
}
template <typename T>
int CreateAclNchTensor(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;
}
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 = {1, 1, 6, 6};
std::vector<int64_t> roisShape = {1, 4};
std::vector<int64_t> batchIndicesShape = {1};
std::vector<int64_t> outShape = {1, 1, 3, 3};
void* selfDeviceAddr = nullptr;
void* roisDeviceAddr = nullptr;
void* batchIndicesDeviceAddr = nullptr;
void* outDeviceAddr = nullptr;
aclTensor* self = nullptr;
aclTensor* rois = nullptr;
aclTensor* batchIndices = nullptr;
aclTensor* out = nullptr;
std::vector<float> selfHostData = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36};
std::vector<float> roisHostData = {-2.0, -2.0, 22.0, 22.0};
std::vector<int32_t> batchIndicesHostData = {0};
std::vector<float> outHostData = {4.5, 6.5, 8.5, 16.5, 18.5, 20.5, 28.5, 30.5, 32.5};
// 创建self aclTensor
ret = CreateAclNchTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
CHECK_RET(ret == ACL_SUCCESS, return ret);
// 创建rois aclTensor
ret = CreateAclTensor(roisHostData, roisShape, &roisDeviceAddr, aclDataType::ACL_FLOAT, &rois);
CHECK_RET(ret == ACL_SUCCESS, return ret);
// 创建batchIndices aclTensor
ret = CreateAclTensor(batchIndicesHostData, batchIndicesShape, &batchIndicesDeviceAddr, aclDataType::ACL_INT32, &batchIndices);
CHECK_RET(ret == ACL_SUCCESS, return ret);
// 创建out aclTensor
ret = CreateAclNchTensor(outHostData, outShape, &outDeviceAddr, aclDataType::ACL_FLOAT, &out);
CHECK_RET(ret == ACL_SUCCESS, return ret);
const char* mode = "avg";
int outputHeight = 3;
int outputWidth = 3;
int samplingRatio = 0;
float spatialScale = 1.0f;
// 3. 调用CANN算子库API
uint64_t workspaceSize = 0;
aclOpExecutor* executor;
// 调用aclnnRoiAlign第一段接口
ret = aclnnRoiAlignGetWorkspaceSize(self, rois, batchIndices, mode, outputHeight, outputWidth, samplingRatio, spatialScale, out, &workspaceSize, &executor);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnRoiAlignGetWorkspaceSize 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aclnnRoiAlign第二段接口
ret = aclnnRoiAlign(workspaceAddr, workspaceSize, executor, stream);
CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnRoiAlign 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 resultData 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,需要根据具体API的接口定义修改
aclDestroyTensor(self);
aclDestroyTensor(rois);
aclDestroyTensor(batchIndices);
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(roisDeviceAddr);
aclrtFree(batchIndicesDeviceAddr);
aclrtFree(outDeviceAddr);
if (workspaceSize > 0) {
aclrtFree(workspaceAddr);
}
aclrtDestroyStream(stream);
aclrtResetDevice(deviceId);
aclFinalize();
return 0;
}