pip和编译安装TensorFlow和OpenCV

我会带着你远行 2022-04-17 06:26 259阅读 0赞

### CPU下利用pip安装和卸载Tensorflow ###

卸载TensorFlow

pip uninstall tensorflow

安装指定版本的TensorFlow

pip install tensorflow==1.10

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70][]

pip安装opencv

pip install opencv-python

[ubuntu系统下编译安装TensorFlow][ubuntu_TensorFlow]

[ICU的安装][ICU]

[ICU58.1的下载][ICU58.1]

[ICU52.1的下载][ICU52.1]

sudo mkdir /usr/local/icu

cd icu/source

./configure --prefix=/usr/local/icu

make

sudo make install

配置环境变量：

sudo gedit /etc/profile

在末尾添加

export LD\_LIBRARY\_PATH=/usr/local/icu/lib:$LD\_LIBRARY\_PATH

export PATH=/usr/local/icu/bin:$PATH

source /etc/profile

第0步：安装java

[ubuntu下配置Java环境][ubuntu_Java]

以上配置完java环境变量后，重启一下电脑。

Bazel在编译TensorFlow时，否则会报错。

第1步：安装[Bazel][]

1.1 下载安装[Bazel在ubuntu下的安装][Bazel_ubuntu]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 1][]

1.2 下载下来的bazel.sh文件 是没有执行权限的，所以要改变权限设置

chmod +x bazel-<version>-installer-linux-x86_64.sh

查看一下bazel.sh文件的权限，是否有执行权限：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 2][]

1.3 在终端执行以下命令，

./bazel-<version>-installer-linux-x86_64.sh --user

会将bazel安装在 $HOME/bin下，如下图：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 3][]

1.4 配置 bazel的环境变量，

gedit $HOME/.bashrc

在文件末尾加入

export PATH=$HOME/bin:$PATH

如下图所示：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 4][]

1.5 环境变量生效

source $HOME/.bashrc

第2步：下载 [NCCL][]，用来支持 多GPU

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 5][]

下载后解压：

tar -xf nccl_2.3.7-1+cuda10.0_x86_64.txz

复制到/usr/local/下

sudo cp -R nccl_2.3.7-1+cuda10.0_x86_64 /usr/local

在/usr/local下多了一个 nccl\_2.3.7-1+cuda10.0\_x86\_64的

sudo ldconfig

配置TensorFlow的编译选项

./configure

利用Bazel编译TensorFlow

bazel build --config=opt --config=cuda //tensorflow/tools/pip_package:build_pip_package

第3步:下载TensorFlow

首先查看一下 命令 git clone的帮助文档，如下：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 6][]

然后利用 git clone命令下载tensorflow

git clone https://github.com/tensorflow/tensorflow

git clone --recurse-submodules https://github.com/tensorflow/tensorflow

检出r1.12版本

git checkout r1.12

[编译安装TensorFlow的官方文][TensorFlow]

[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70]: /images/20220417/902e8681e6dd4baf947fe0a80b5a631b.png
[ubuntu_TensorFlow]: https://www.python36.com/how-to-install-tensorflow-gpu-with-cuda-10-0-for-python-on-ubuntu/
[ICU]: http://site.icu-project.org/download
[ICU58.1]: http://download.icu-project.org/files/icu4c/58.1/
[ICU52.1]: http://download.icu-project.org/files/icu4c/52.1/
[ubuntu_Java]: https://blog.csdn.net/zhangping1987/article/details/39644063
[Bazel]: https://bazel.build/
[Bazel_ubuntu]: https://docs.bazel.build/versions/master/install-ubuntu.html
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 1]: /images/20220417/cdb9868824cb47ca847f3b4ac51ae061.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 2]: /images/20220417/2c3ef81eee104f2eaa073aa027406fc9.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 3]: /images/20220417/60152c69c8054e29b5d5d6af2b307bbd.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 4]: /images/20220417/6ba81c6b66d44304ac0e1706a6218134.png
[NCCL]: https://developer.nvidia.com/nccl/nccl-download
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 5]: /images/20220417/bc8a209bf82143f2bfd551480d92319a.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3d4Y2R6aGFuZ3Bpbmc_size_16_color_FFFFFF_t_70 6]: /images/20220417/74a0925a519f4300ae61e7c508ec8d24.png
[TensorFlow]: https://tensorflow.google.cn/install/source