先给出最终编译好的文件：链接:https://pan.baidu.com/s/130VQjBT6JXJl0yAOOKJJkw 密码:arf0
tensorflow-1.12.3-cp36-cp36m-linux_x86_64.whl
链接:https://pan.baidu.com/s/1V3bNFFClMELeD7ExIvRIsQ 密码:oycm
tensorflow-1.14.0-cp36-cp36m-linux_x86_64.whl
ubuntu18.04 86位系统 tensorflow1.12 python 3.6 在docker中编译的结果
整个编译时间么。。差不多两小时把 = =
tf1.14配套的版本
tf 1.14，bazel 0.24.1，gcc,g++ 4.8
1\apt install git
2\降低gcc版本，看前面一篇文章 git clone https://github.com/tensorflow/tensorflow.git
3\然后安装bazel 这里就不装java了 https://docs.bazel.build/versions/master/install-ubuntu.html
4\https://www.tensorflow.org/install/source?hl=zh-cn 按照这个文档操作，一定注意对应的版本问题
tf 1.12 gcc 4.8 bazel 0.15这些都是测试通过的，如果自己用高版本很可能碰到意想不到的错误
5\卸载掉原先掉tensorflow pip uninstall tensorflow
6\去bazel网站下载0.15的bazel https://github.com/bazelbuild/bazel/releases?after=0.19.0
下载好之后，另外打开一个窗口，通过docker cp的方式 src dst这样 38f30b3590fb这个是打开的容器id
docker cp /Users/zakj/Downloads/bazel-0.15.0-installer-linux-x86_64.sh 38f30b3590fb:/
7\然后根据这个文档安装bazel https://docs.bazel.build/versions/master/install-ubuntu.html
8\这里是很关键的一步骤，git checkout branch_name # r1.9, r1.10, etc. 一定要ckeckout 到r1.12
否则会提示bazel需要0.24 这时候gcc版本又不对了，又没人测试过，就完蛋了。这时候还要注意g++的版本
https://www.tensorflow.org/install/source?hl=zh-cn
9\接下来的操作很容易了，bazel build -c opt --copt=-mavx --copt=-mavx2 --copt=-mfma --copt=-msse4.1 --copt=-msse4.2 --copt=-mfpmath=both -k //tensorflow/tools/pip_package:build_pip_package
10\不幸的是，又遇到了错误，错误原因好像是g++版本不对，修改了和gcc对应的4.8的g++版本，又对了
https://www.linuxidc.com/Linux/2016-08/134546.htm
终于大功告成
docker 使用手册 https://yeasy.gitbooks.io/docker_practice/introduction/what.html
/root/.cache/bazel/_bazel_root/68a62076e91007a7908bc42a32e4cff9/external/com_google_absl/absl/base/BUILD.bazel:118:1: C++ compilation of rule '@com_google_absl//absl/base:base' failed (Exit 1)
通过这个指令，下面这张图我也不知道哪里来的
不知道哪里来的
删除原来的镜像 我好像只能先停止再删除
docker container prune -f docker image rm imageid
https://yeasy.gitbooks.io/docker_practice/image/rm.html
https://colobu.com/2018/05/15/Stop-and-remove-all-docker-containers-and-images/
好吧还是报错来
ERROR: /root/.cache/bazel/_bazel_root/68a62076e91007a7908bc42a32e4cff9/external/com_google_absl/absl/base/BUILD.bazel:118:1: C++ compilation of rule '@com_google_absl//absl/base:base' failed (Exit 1)
gcc: error trying to exec 'cc1plus': execvp: No such file or directory
Target //tensorflow/tools/pip_package:build_pip_package failed to build
Use --verbose_failures to see the command lines of failed build steps.
INFO: Elapsed time: 344.642s, Critical Path: 0.13s
INFO: 0 processes.
ERROR: /root/.cache/bazel/_bazel_root/68a62076e91007a7908bc42a32e4cff9/external/aws/BUILD.bazel:12:1: C++ compilation of rule '@aws//:aws' failed (Exit 4)
解决参考：
1、https://github.com/tensorflow/serving/issues/1245
2、前面几篇文章
3、https://www.tensorflow.org/install/docker
https://stackoverflow.com/questions/34183997/why-am-i-getting-docker-error-c-compilation-of-rule-tensorflow-corekernelhttps://github.com/tensorflow/tensorflow/issues/20677https://blog.csdn.net/lanbeiliu/article/details/90293642 内存不足http://m.sohu.com/a/304842717_99979179 从CPU指令集优化
#Install bazel
#download tensorflow src
#Checkout tagged version
#./configure
#Compile
bazel build -c opt --copt=-march=native //tensorflow/tools/pip_package:build_pip_package
#bazel build -c opt --copt=-mavx --copt=-msse4.1 --copt=-msse4.2 //tensorflow/tools/pip_package:build_pip_package
#Build the PIP package
bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/tensorflow_pkg
#uninstall previous tensorflow version (if applicable)
#Install PIP package
pip install /tmp/tensorflow_pkg/tensorflow-1.2.0rc2-cp27-cp27m-macosx_10_12_x86_64.whl
#References
#https://www.tensorflow.org/install/install_sources
#https://stackoverflow.com/questions/41293077/how-to-compile-tensorflow-with-sse4-2-and-avx-instructions
#https://en.wikipedia.org/wiki/Advanced_Vector_Extensions#AVX2
#https://en.wikipedia.org/wiki/SSE4
#https://en.wikipedia.org/wiki/FMA_instruction_set
#https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/x86-Options.html#x86-Options
#https://gcc.gnu.org/onlinedocs/gcc-4.5.3/gcc/i386-and-x86_002d64-Options.html
网友评论