系统与环境要求 Windows10系统 GTX1050Ti CUDA10.x VS2017 TensorRT7.0.0.11
01安装与配置 下载路径: https://developer.nvidia.com/TensorRT 首先需要下载TensorRT的ZIP格式文件到本地,然后解压缩到 D:\TensorRT-7.0.0.11
然后打开VS2017,新建一个空项目,分别配置
包含目录 D:\TensorRT-7.0.0.11\include 库目录 D:\TensorRT-7.0.0.11\lib 链接器![五分钟搞定VS2017+TensorRT环境搭建 五分钟搞定VS2017+TensorRT环境搭建](https://cms-pic.yhzz.com.cn/1683799492724.jpg)
![五分钟搞定VS2017+TensorRT环境搭建1 五分钟搞定VS2017+TensorRT环境搭建1](https://cms-pic.yhzz.com.cn/1683799492891.jpg)
重启VS即可。
02代码验证与测试 2020年初,我写过的pytorch程序有个Hello Wrold的版本的模型就是mnist.onnx,我来测试一下是否可以通过TensorRT来实现对ONNX格式模型加载。重启VS2017之后在原来的空项目上然后添加一个cpp文件,把下面的代码copy到cpp文件中: include include include
include “NvInfer.h” include “NvOnnxParser.h”
using namespace nvinfer1; using namespace nvonnxparser;
class Logger : public ILogger { void log(Severity severity, const char* msg) override { // suppress info-level messages if (severity != Severity::kINFO) std::cout << msg << std::endl; } } gLogger;
int main(int argc, char* argv) {
std::string onnx_filename = “D:/python/pytorch_tutorial/cnn_mnist.onnx”;
IBuilder builder = createInferBuilder(gLogger);
nvinfer1::INetworkDefinition network = builder->createNetworkV2(1U << static_cast(NetworkDefinitionCreationFlag::kEXPLICIT_BATCH));
auto parser = nvonnxparser::createParser(network, gLogger);
parser->parseFromFile(onnx_filename.c_str(), 2);
for (int i = 0; i < parser->getNbErrors(); ++i)
{
std::cout << parser->getError(i)->desc() << std::endl;
}
printf(“tensorRT load onnx mnist model…\n”);
return 0;
}
编译运行直接运行输出:
恭喜你!TensorRT在Windows10下开发环境配置成功了!绝对在5分钟内搞定,前提是先预装好前面说的那些依赖软件与相关的库!