Tensorflow 编译吐血历史
Table of Contents

关于

因为要在CentOS6中运行TensorFlow,但是TensorFlow依赖的Glibc版本都比较高,所以只能从源码编译,蛋疼!

因为编译TensorFlow需要bazel,而bazel官方编译好的包也需要CentOS7,所以bazel也需要从源码编译!

bazel 需要Java8,所以如果没有安装Java8的还需要先安装Java8

安装Java8

  1. 去Oracle官网下一个:https://www.oracle.com/technetwork/java/javase/downloads/jdk8-downloads-2133151.html
    或者用我的百度云镜像 https://pan.baidu.com/s/161G0d3GYsw5dY6En4SHpjA
  2. 下载完解压即可,然后配置JAVA_HOME环境变量,鉴于你的系统可能有多个JAVA版本,建议你只在脚本中或者在命令行中临时设置这个变量
export JAVA_HOME=path-to-java-directory
export PATH=path-to-java-directory/bin:$PATH

我的做法是将上述保存到env.sh中,然后在 compile.sh 中加载 source env.sh

通过 java -version 确认当前使用的是刚下载的java8,而不是系统的java7等版本。

编译Bazel

参考:https://docs.bazel.build/versions/master/install-compile-source.html

设置好JAVA8环境变量之后, GCC版本也不能太低,安装高版本GCC6.0

执行编译命令 ./compile.sh,然后发现错误:

bazel-out/host/bin/third_party/protobuf/3.4.0/js_embed: /usr/lib64/libstdc++.so.6: version `GLIBCXX_3.4.21' not found (required by bazel-out/host/bin/third_party/protobuf/3.4.0/js_embed)

参考文章https://www.jianshu.com/p/6f9652a9410d将libstdc++改为静态链接

tools/cpp/unix_cc_configure.bzl:222
  "-lstdc++"     改为          "-l:libstdc++.a"

编译TensorFlow

  1. 运行配置./configure
  2. 编译pip包 bazel build --config=opt //tensorflow/tools/pip_package:build_pip_package

配置的时候,OPENCL关闭,jemalloc也要关闭,将所有都关闭,哈哈

编译时增加 -lrt 参数,参考 https://www.cnblogs.com/lasclocker/p/9578623.html

运行测试

将上述编译好的三个文件 libtensorflow_framework.so, libtensorflow.jar, libtensorflow_jni.so 复制到一个目录中,然后创建测试代码 HelloTF.java

import org.tensorflow.Graph;
import org.tensorflow.Session;
import org.tensorflow.Tensor;
import org.tensorflow.TensorFlow;

public class HelloTF {
  public static void main(String[] args) throws Exception {
    try (Graph g = new Graph()) {
      final String value = "Hello from " + TensorFlow.version();

      // Construct the computation graph with a single operation, a constant
      // named "MyConst" with a value "value".
      try (Tensor t = Tensor.create(value.getBytes("UTF-8"))) {
        // The Java API doesn't yet include convenience functions for adding operations.
        g.opBuilder("Const", "MyConst").setAttr("dtype", t.dataType()).setAttr("value", t).build();
      }

      // Execute the "MyConst" operation in a Session.
      try (Session s = new Session(g);
           Tensor output = s.runner().fetch("MyConst").run().get(0)) {
        System.out.println(new String(output.bytesValue(), "UTF-8"));
      }
    }
  }
}

编译 javac -cp libtensorflow.jar HelloTF.java

执行 java -cp libtensorflow.jar:. -Djava.library.path=. HelloTF 如果加载so失败,可以增加参数 -Dorg.tensorflow.NativeLibrary.DEBUG=1

created in 2018-11-07