tensorflow学习笔记2：c++程序静态链接tensorflow库加载模型文件

2024年4月18日 56次阅读 来源: tensorflow

首先需要搞定tensorflow c++库，搜了一遍没有找到现成的包，于是下载tensorflow的源码开始编译；

tensorflow的contrib中有一个makefile项目，极大的简化的接下来的工作；

按照tensorflow makefile的说明文档，开始做c++库的编译：

 

1. 下载依赖

在tensorflow的项目顶层运行：

tensorflow/contrib/makefile/download_dependencies.sh

东西会下载到tensorflow/contrib/makefile/downloads/目录里；

 

2. 在linux下进行编译

首先确保编译工具都已经装好了：

sudo apt-get install autoconf automake libtool curl make g++ unzip zlib1g-dev git python

然后运行编译脚本；

注意：运行之前打开看一眼，第一步竟然是把tensorflow/contrib/makefile/downloads/目录里的东西清空然后重新下载。。。注掉注掉

tensorflow/contrib/makefile/build_all_linux.sh

然后在tensorflow/contrib/makefile/gen/lib/libtensorflow-core.a就看到静态库了；

 

3. 准备好加载模型的c++代码

#include "tensorflow/core/public/session.h"
#include "tensorflow/core/platform/env.h"

using namespace tensorflow;

int main(int argc, char* argv[]) {
  // Initialize a tensorflow session
  Session* session;
  Status status = NewSession(SessionOptions(), &session);
  if (!status.ok()) {
    std::cout << status.ToString() << "\n";
    return 1;
  }

  // Read in the protobuf graph we exported
  // (The path seems to be relative to the cwd. Keep this in mind
  // when using `bazel run` since the cwd isn't where you call
  // `bazel run` but from inside a temp folder.)
  GraphDef graph_def;
  status = ReadBinaryProto(Env::Default(), "models/test_graph.pb", &graph_def);
  if (!status.ok()) {
    std::cout << status.ToString() << "\n";
    return 1;
  }

  // Add the graph to the session
  status = session->Create(graph_def);
  if (!status.ok()) {
    std::cout << status.ToString() << "\n";
    return 1;
  }

  // Setup inputs and outputs:

  // Our graph doesn't require any inputs, since it specifies default values,
  // but we'll change an input to demonstrate.
  Tensor a(DT_FLOAT, TensorShape());
  a.scalar<float>()() = 3.0;

  Tensor b(DT_FLOAT, TensorShape());
  b.scalar<float>()() = 2.0;

  Tensor x(DT_FLOAT,TensorShape());
  x.scalar<float>()() = 10.0;

  std::vector<std::pair<string, tensorflow::Tensor>> inputs = {
    { "a", a },
    { "b", b },
    { "x", x },
  };

  // The session will initialize the outputs
  std::vector<tensorflow::Tensor> outputs;

  // Run the session, evaluating our "y" operation from the graph
  status = session->Run(inputs, {"y"}, {}, &outputs);
  if (!status.ok()) {
    std::cout << status.ToString() << "\n";
    return 1;
  }

 // Grab the first output (we only evaluated one graph node: "c")
  // and convert the node to a scalar representation.
  auto output_y = outputs[0].scalar<float>();

  // (There are similar methods for vectors and matrices here:
  // https://github.com/tensorflow/tensorflow/blob/master/tensorflow/core/public/tensor.h)

  // Print the results
  std::cout << outputs[0].DebugString() << "\n"; // Tensor<type: float shape: [] values: 32>
  std::cout << output_y() << "\n"; // 32

  // Free any resources used by the session
  session->Close();
  return 0;
}

保存成load_graph.cc；

 

写Makefile：

TARGET_NAME := load_graph

TENSORFLOW_MAKEFILE_DIR := /mnt/data/tensorflow/tensorflow/contrib/makefile

INCLUDES := \
-I /usr/local/lib/python3.6/dist-packages/tensorflow/include

NSYNC_LIB := \
$(TENSORFLOW_MAKEFILE_DIR)/downloads/nsync/builds/default.linux.c++11/nsync.a

PROTOBUF_LIB := \
$(TENSORFLOW_MAKEFILE_DIR)/gen/protobuf/lib/libprotobuf.a

TENSORFLOW_CORE_LIB := \
-Wl,--whole-archive $(TENSORFLOW_MAKEFILE_DIR)/gen/lib/libtensorflow-core.a -Wl,--no-whole-archive

LIBS := \
$(TENSORFLOW_CORE_LIB) \
$(NSYNC_LIB) \
$(PROTOBUF_LIB) \
-lpthread \
-ldl

SOURCES := \
load_graph.cc

$(TARGET_NAME):
	g++ -std=c++11 $(SOURCES) $(INCLUDES) -o $(TARGET_NAME) $(LIBS)

clean:
	rm $(TARGET_NAME)

这里的tensorflow-core、nsync和protobuf全都用静态链接了，这些静态库以后考虑都放一份到系统目录下；

 

有几个点需要注意：

1) INCLUDE使用了python3.6的带的tensorflow头文件，只是觉得反正python都已经带头文件了，就不需要再另外拷一份头文件进系统目录了；

2) nsync库是多平台的，因而可能需要仔细分析一下nsync的编译结果所在位置，尤其如果是交叉编译的话；

3) 链接顺序不能错，tensorflow-core肯定要在其它两个前面；

4) tensorflow_core库需要全链接进来，否则会出现这个错：tensorflow/core/common_runtime/session.cc:69] Not found: No session factory registered for the given session options: {target: “” config: } Registered factories are {}.

    想想也大概能知道为什么，肯定是在静态代码层面只依赖父类，然后再在运行时通过名字找子类，所以在符号层面是不直接依赖子类的，不强制whole-archive的话，子类一个都带不进来；

 

4. 运行程序

运行前先看看事先准备好的graph在不在预定位置，生成graph的方法见上一篇；

运行一下，没啥好说的，结果正确。

 

参考：

http://blog.163.com/wujiaxing009@126/blog/static/7198839920174125748893/

https://blog.csdn.net/xinchen1234/article/details/78750079

    原文作者：tensorflow
    原文地址: https://www.cnblogs.com/ZisZ/p/9145164.html
    本文转自网络文章，转载此文章仅为分享知识，如有侵权，请联系博主进行删除。
点赞