Play with TensorFlow - GerryZhang0925/dev_env GitHub Wiki

Table of Contents 1. install tensorflow 1.1 install for python 1.2 install for anaconda 2. Run TensorFlow 2.1 Hello World 3. Rebuild Tensorflow (GPU) on CUDA 10.0 and cuDNN 7.3 3.1 Install CUDA 10.0 3.2 Configure CUDA Path 3.3 Build CUDA samples 3.4 Install cuDNN 7.3 3.5 Build cuDNN samples 4.0 Install Bazel 0.18.1 5.0 Build TensorFlow 5.1 Preparation for common environment 5.2 Preparation for virtual environment 5.3 Download and build 5.4 Installation

With the simple pip installation, override existing packages will be overridded.

   > pip install tensorflow

As a solution, a virtual environment can be used.

   > pip install virtualenv
   > cd ~
   > mkdir envs
   > virtualenv ~/envs/tensorflow
   > source ~/envs/tensorflow/bin/activate
   (tensorflow)$ pip install tensorflow  # pip install tensorflow-gpu for the GPU-enabled version
   (tensorflow)$ deactivate       # exit the virtual environment

To use tensorflow often, the following command can be appended to ~/.bashrc file:

   alias tensorflow="source ~/envs/tensorflow/bin/activate"

Create virtual environment, and install tensorflow on it.

   > conda create --name tensorflow_env3.6 python=3.6 anaconda
   > source activate tensorflow_env3.6
   (tensorflow)$ conda install -c anaconda tensorflow-gpu
   (tensorflow)$ source deactivate tensorflow_env3.6

Import the tensorflow package.

   (tensorflow)$ python
   >>> import tensorflow as tf
   >>> print(tf.__version__)
   1.12.0

   sudo dpkg -i cuda-repo-ubuntu1604-10-0-local-10.0.130-410.48_1.0-1_amd64.deb
   sudo apt-key add /var/cuda-repo-10–0-local-10.0.130–410.48/7fa2af80.pub
   sudo apt-get update
   sudo apt-get install cuda

Edit ~/.bashrc to append following lines.

   export PATH=/usr/local/cuda-10.0/bin${PATH:+:${PATH}}
   export LD_LIBRARY_PATH=/usr/local/cuda-10.0/lib64:/usr/lib/x86_64-linux-gnu/:/usr/local/cuda-10.0/targets/x86_64-linux/lib:${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

Goto /usr/local/cuda-10.0/samples and run

   sudo make

Download cuDNN7.3 from https://developer.nvidia.com/rdp/cudnn-archive and run following commands.

   sudo dpkg -i libcudnn7_7.3.0.29–1+cuda10.0_amd64.deb
   sudo dpkg -i libcudnn7-dev_7.3.0.29–1+cuda10.0_amd64.deb
   sudo dpkg -i libcudnn7-doc_7.3.0.29–1+cuda10.0_amd64.deb

Run following instructions.

   cp -r /usr/src/cudnn_samples_v7/ ~
   cd ~/cudnn_samples_v7/mnistCUDNN
   make clean && make
   ./mnistCUDNN

Bazel is a tool that will be used for TensorFlow building. Download bazel-0.18.1-installer-linux-x86_64.sh from https://github.com/bazelbuild/bazel/releases and install it as following.

   chmod +x bazel-0.18.1-installer-linux-x86_64.sh
   ./bazel-0.18.1-installer-linux-x86_64.sh --user
   sudo apt install curl
   sudo apt-get install openjdk-8-jdk
   sudo add-apt-repository ppa:webupd8team/java
   sudo apt-get update && sudo apt-get install oracle-java8-installer
   echo "deb [arch=amd64] http://storage.googleapis.com/bazel-apt stable jdk1.8" | sudo tee /etc/apt/sources.list.d/bazel.list
   curl https://bazel.build/bazel-release.pub.gpg | sudo apt-key add -
   sudo apt-get update
   sudo apt-get install bazel

Install python3-distutils as following.

   sudo apt-get install python3-distutils
   sudo apt install python-dev python-pip

   pip install -U pip six numpy wheel mock
   pip install -U keras_applications==1.0.5 --no-deps
   pip install -U keras_preprocessing==1.0.3 --no-deps
   pip install h5py==2.8.0
   pip install --upgrade pip setuptools
   conda install libgcc

   git clone https://github.com/tensorflow/tensorflow.git
   cd tensorflow
   git checkout r1.12
   # All tests (for C++ changes).
   bazel test //tensorflow/...
   # All Python tests (for Python front-end changes).
   bazel test //tensorflow/python/...
   # All tests (with GPU support).
   bazel test -c opt --config=cuda //tensorflow/...
   bazel test -c opt --config=cuda //tensorflow/python/...

It was processing about an hour on my machine. As a result I have about 60 failed test results but it doesn’t impact build process. Then configure the tensorflow.

   ./configure
   Do you wish to build TensorFlow with CUDA support? [y/N]: y
   Please specify the CUDA SDK version you want to use. [Leave empty to default to CUDA 9.0]: 10
   Please specify the NCCL version you want to use. \
   If NCCL 2.2 is not installed, then you can use version 1.3 that can be fetched\
   automatically but it may have worse performance with multiple GPUs. [Default is 2.2]: 1.3
   time bazel build --verbose_failures --config=opt --config=cuda //tensorflow/tools/pip_package:build_pip_package
   ./bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/tensorflow_pkg --project_name tensorflow_gpu_cuda_10.0

   pip install /tmp/tensorflow_pkg/tensorflow_gpu_cuda_10.0-1.12.0-cp36-cp36m-linux_x86_64.whl