ubuntu_tensorflowGPU_CPU_without_AVX_wheel

Build tensorflow-gpu 1.12.0 wheel (Ubuntu 18.10) which CPU does not support AVX (Intel Pentium GOLD G5400)

It took me around 6 hours to successfully build this wheel !

some details of my environment:

Intel® Pentium® Gold G5400 Processor 4M Cache, 3.70 GHz
GeForce GTX 1060/ 6GB
Ubuntu 18.10
tensorflow-gpu 1.12.0
CUDA 10.0
cuDNN 7
The latest version of Anaconda with python 3.7.1

Building Steps:

1. 
   

    sudo apt-get update
    sudo apt-get upgrade
   

2. GPU supports CUDA
   

    lspci | grep -i nvidia
   

3. install dependencies and repository
   

    sudo apt-get install build-essential
    sudo apt-get install cmake git unzip zip
    sudo add-apt-repository ppa:deadsnakes/ppa
    sudo apt-get update
    python -> I use Anaconda ENV
   

4. install Nvidia CUDA 10.0
   

    I use the way of cuda_10.0.130_410.48_linux.run for installation
   

5. cofigure your building environments
   

    echo 'export PATH=/usr/local/cuda-10.0/bin${PATH:+:${PATH}}' >> ~/.bashrc
    echo 'export LD_LIBRARY_PATH=/usr/local/cuda-10.0/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}' >> ~/.bashrc
    source ~/.bashrc
    sudo ldconfig
    nvidia-smi (if you concern...)
   

6. install cuDNN 7 with the lazy way
   

    libcudnn7_7.4.1.5-1+cuda10.0_amd64.deb
    libcudnn7-dev_7.4.1.5-1+cuda10.0_amd64.deb
    libcudnn7-doc_7.4.1.5-1+cuda10.0_amd64.deb
    sudo apt-get install libcupti-dev
    echo 'export LD_LIBRARY_PATH=/usr/local/cuda/extras/CUPTI/lib64:$LD_LIBRARY_PATH' >> ~/.bashrc
    (in Anaconda ENV) sudo apt-get install python3-numpy python3-dev python3-pip python3-wheel
   

7. official building style of tensorflow
   

   wget https://github.com/bazelbuild/bazel/releases/download/0.2X.0/bazel-0.2X.0-installer-linux-x86_64.sh
   chmod +x bazel-0.2X.0-installer-linux-x86_64.sh
   ./bazel-0.2X.0-installer-linux-x86_64.sh --user
   echo 'export PATH="$PATH:$HOME/bin"' >> ~/.bashrc
   source ~/.bashrc
   git clone https://github.com/tensorflow/tensorflow.git
   cd tensorflow
   ./configure
   

8. selections of configurations (follow your heart !)
   

    You have bazel 0.17.2 installed.
    Please specify the location of python. [Default is /usr/bin/python]: (Anaconda-env-path)
    
    Found possible Python library paths:
    /...
    /...
    Please input the desired Python library path to use.  Default is [/usr/local/lib/python3.5/dist-packages]
   
    Do you wish to build TensorFlow with Apache Ignite support? [Y/n]: n
   
    Do you wish to build TensorFlow with XLA JIT support? [Y/n]: n
   
    Do you wish to build TensorFlow with OpenCL SYCL support? [y/N]: n
   
    Do you wish to build TensorFlow with ROCm support? [y/N]: n
   
    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]:  10.0
   
    Please specify the location where CUDA 10.0 toolkit is installed. Refer to README.md for more details. [Default is /usr/local/cuda]: 
    
    Please specify the cuDNN version you want to use. [Leave empty to default to cuDNN 7]: 
    
    Please specify the location where cuDNN 7 library is installed. Refer to README.md for more details. [Default is /usr/local/cuda]: 
   
    Do you wish to build TensorFlow with TensorRT support? [y/N]: n
    
    Please specify the locally installed NCCL version you want to use. [Default is to use https://github.com/nvidia/nccl]: 
   
    Please note that each additional compute capability significantly increases your build time and binary size. [Default is: 6.1]: 
    
    Do you want to use clang as CUDA compiler? [y/N]: n
    	
    Please specify which gcc should be used by nvcc as the host compiler. [Default is /usr/bin/gcc]:  gcc-7
    
    Do you wish to build TensorFlow with MPI support? [y/N]: n
    
    Please specify optimization flags to use during compilation when bazel option "--config=opt" is specified [Default is -march=native]: 
    
    Would you like to interactively configure ./WORKSPACE for Android builds? [y/N]: n
    ...
    Configuration finished
   

9. build pip package (it will take around 2.5~3.5 hours!)
   

    bazel build --config=opt //tensorflow/tools/pip_package:build_pip_package
    ...
    INFO: Build completed successfully, 8036 total actions
   

10. build pip package
    

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

11. build the invincible WHL file (you should prey for your GOD now)

    bazel-bin/tensorflow/tools/pip_package/build_pip_package tensorflow_pkg
    

12. find the invincible WHL file and install it ! (you can drink your whisky now)

    ...
    ...
    CST 2018 : === Output wheel file is in:
    /.../tensorflow/tensorflow_pkg
    

pip install tensorflow-1.12.0-cp36-cp36m-linux_x86_64.whl

Find your favorite tensorflow scripts for testing this whl~~~

^_^

Author

kao.kuntai@GMAIL

License

This WHL is totally free for the installation of tensorflow-gpu which has no AVX support of Intel CPU serious.