Skip to content

tensorflow_cc-examples.md

Latest commit

 

History

History
166 lines (135 loc) · 4.56 KB

tensorflow_cc-examples.md

File metadata and controls

166 lines (135 loc) · 4.56 KB

TensorFlow C++ 示例代码

环境

  • 系统：Ubuntu
  • 版本：16.04.3 LTS
  • 处理器：Intel(R) Core(TM) i7-7700K CPU @ 4.20GHz
  • 内存：16.0GB
  • 类型：64位操作系统 64位处理器
  • 显卡：索泰GTX1060 6G

写在前面

使用 C++ 语言来编写 TensorFlow 程序与使用 Python 语言一样，需要安装 TensorFlow 环境。参照将 Tensorflow 源码编译成 C++ 库文件这篇教程的安装步骤可以完成 TensorFlow C++ 环境的部署。

但是 C++ 源码需要编译后才能执行，编译所使用的 makefile 如下：

target = tfcc_test
cc = g++ -std=c++11
include = -I/usr/local/tensorflow/include
lib = -L/usr/local/tensorflow/lib -ltensorflow_framework -ltensorflow_cc
flag = -Wl,-rpath=/usr/local/tensorflow/lib
source = ./src/main.cc

$(target): $(source)
	$(cc) $(source) -o $(target) $(include) $(lib) $(flag)

clean:
	rm $(target)

run:
	./$(target)

示例代码

  • 创建 Session

    #include "tensorflow/cc/client/client_session.h"

using namespace tensorflow;

int main()
{
    auto root = Scope::NewRootScope();
    auto p_session = new ClientSession(root);
    delete p_session;
    return 0;
}

  • 常量

    #include "tensorflow/cc/client/client_session.h"
#include "tensorflow/cc/ops/standard_ops.h"

using namespace tensorflow;
using namespace tensorflow::ops;
using namespace std;

int main()
{
    auto root = Scope::NewRootScope();
    auto w = Const(root, 2, {});
    auto p_session = new ClientSession(root);
    vector<Tensor> outputs;
    p_session->Run({w}, &outputs);
    LOG(INFO) << "w = " << outputs[0].scalar<int>();
    delete p_session;
    return 0;
}

  • 变量

    #include "tensorflow/cc/client/client_session.h"
#include "tensorflow/cc/ops/standard_ops.h"

using namespace tensorflow;
using namespace tensorflow::ops;
using namespace std;

int main()
{
    auto root = Scope::NewRootScope();
    auto x = Variable(root, {}, DataType::DT_INT32);
    auto assign_x = Assign(root, x, 3); // initializer for x
    auto y = Variable(root, {2, 3}, DataType::DT_FLOAT);
    auto assign_y = Assign(root, y, RandomNormal(root, {2, 3}, DataType::DT_FLOAT)); // initializer for y
    auto p_session = new ClientSession(root);
    p_session->Run({assign_x, assign_y}, nullptr); // initialize
    vector<Tensor> outputs;
    p_session->Run({x, y}, &outputs);
    LOG(INFO) << "x = " << outputs[0].scalar<int>();
    LOG(INFO) << "y = " << outputs[1].matrix<float>();
    delete p_session;
    return 0;
}

  • 矩阵运算

    #include "tensorflow/cc/client/client_session.h"
#include "tensorflow/cc/ops/standard_ops.h"

using namespace tensorflow;
using namespace tensorflow::ops;
using namespace std;

int main()
{
    auto root = Scope::NewRootScope();
    auto x = Variable(root, {5, 2}, DataType::DT_FLOAT);
    auto assign_x = Assign(root, x, RandomNormal(root, {5, 2}, DataType::DT_FLOAT));
    auto y = Variable(root, {2, 3}, DataType::DT_FLOAT);
    auto assign_y = Assign(root, y, RandomNormal(root, {2, 3}, DataType::DT_FLOAT));
    auto xy = MatMul(root, x, y);
    auto z = Const(root, 2.f, {5, 3});
    auto xyz = Add(root, xy, z);
    auto p_session = new ClientSession(root);
    p_session->Run({assign_x, assign_y}, nullptr);
    vector<Tensor> outputs;
    p_session->Run({x, y, z, xy, xyz}, &outputs);
    LOG(INFO) << "x = " << outputs[0].matrix<float>();
    LOG(INFO) << "y = " << outputs[1].matrix<float>();
    LOG(INFO) << "xy = " << outputs[3].matrix<float>();
    LOG(INFO) << "z = " << outputs[2].matrix<float>();
    LOG(INFO) << "xyz = " << outputs[4].matrix<float>();
    delete p_session;
    return 0;
}

  • Placeholder

    #include "tensorflow/cc/client/client_session.h"
#include "tensorflow/cc/ops/standard_ops.h"

using namespace tensorflow;
using namespace tensorflow::ops;
using namespace std;

int main()
{
    auto root = Scope::NewRootScope();
    auto x = Placeholder(root, DataType::DT_INT32);
    auto w = Const(root, 1, {1, 2});
    auto wx = MatMul(root, x, w);
    auto b = Const(root, 2, {2});
    auto wx_b = Add(root, wx, b);
    auto p_session = new ClientSession(root);
    vector<Tensor> outputs;
    p_session->Run({{x, {{1}, {1}, {1}}}}, {wx, wx_b}, &outputs);
    LOG(INFO) << "wx = " << outputs[0].matrix<int>();
    LOG(INFO) << "wx_b = " << outputs[1].matrix<int>();
    delete p_session;
    return 0;
}