Deep-learning

environment installation for DL

Ubuntu (Mac)

Tips for terminal

theme

sudo apt-get install dconf-cli
wget -O gogh https://git.io/vQgMr && chmod +x gogh && ./gogh && rm gogh
# chalk(22), ocean dark

Oh-my-zsh

apt-get install git zsh curl
sh -c "$(curl -fsSL https://github.com/raw/robbyrussell/oh-my-zsh/master/tools/install.sh)"

vi ~/.zshrc
ZSH_THEME=agnoster
DEFAULT_USER=$USER
# or
prompt_context() {}

chsh -s $(which zsh)  # change shell to zsh for user
if passwd needed:
Changing /etc/pam.d/chsh: from:
auth       required   pam_shells.so
to
auth       sufficient   pam_shells.so

Powerfonts

sudo apt-get install fonts-powerline

quick-look

brew cask install qlcolorcode qlstephen qlmarkdown quicklook-json qlimagesize suspicious-package quicklookase qlvideo

silver searcher

 sudo apt-get install silversearcher-ag
 for mac:
 brew install the_silver_searcher

auto-suggestion

git clone https://github.com/zsh-users/zsh-autosuggestions ${ZSH_CUSTOM:-~/.oh-my-zsh}/plugins/zsh-autosuggestions

plugins=(zsh-autosuggestions) 

highlighting

git clone https://github.com/zsh-users/zsh-syntax-highlighting.git ${ZSH_CUSTOM:-~/.oh-my-zsh}/plugins/zsh-syntax-highlighting
plugins=(zsh-syntax-highlighting)
source ~/.zshrc

autojump

sudo apt-get install autojump

. /usr/share/autojump/autojump.sh
#add this to ~/.zshrc

thefuck

sudo apt update
sudo apt install python3-dev python3-pip python-dev python-pip
sudo pip3 install thefuck

brew install thefuck  # OSX

Others

pip install tldr

sshpass

brew install https://github.com/raw/kadwanev/bigboybrew/master/Library/Formula/sshpass.rb

---

Environment for Deep learning

CUDA and drivers

1. Download CUDA , cuda 9.1[ tensorflow only support cuda 8 ]

2. Remove any other installation (sudo apt-get purge nvidia-cuda* - if you want to install the drivers too, then sudo apt-get purge nvidia-*.)

3. sudo sh cuda7.5.18linux.run --override
   

   If cannot install graphic driver.

   1. if error: It appears that an X server is running:

      enter non-GUI mode (tty1,.eg.)

      sudo service lightdm stop
      

   2. if LC_ALL warning

      echo "export LC_ALL="en_US.UTF-8"" >> ~/.zshrc
      

   3. install Nvidia driver manually

      lspci | grep -i nvidia  # check the gpu device
      

      Download the right driver version

      sudo add-apt-repository ppa:graphics-drivers/ppa
      sudo apt update
      sudo apt-get install nvidia-*

      If failed due to kernel issues, it might due to the wrong gcc version.

      Use 'gcc -v' to check gcc version.

      install the right gcc version needed

      sudo update-alternatives --config gcc  # switch gcc version
      

4. Add path to ~/.zshrc or ~/.bashrc

   echo "export PATH=$PATH:/usr/local/cuda-8.0/bin
   export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-8.0/lib64" >> ~/.zshrc

5. Verify CUDA installation

   nvcc --version 
   nvidia-smi

CUDNN

1. Download cudnn:

   1. cudnn8
   2. cudnn9

2. Unzip the cuDNN package.

   tar -xvf cudnn-9.0-linux-x64-v7.tar

3. Copy the following files into the CUDA Toolkit directory.

   sudo cp cuda/include/cudnn.h /usr/local/cuda/include
   sudo cp cuda/lib64/libcudnn* /usr/local/cuda/lib64
   sudo chmod a+r /usr/local/cuda/include/cudnn.h /usr/local/cuda/lib64/libcudnn*

Anoconda

1. Download the installer:

   • Anaconda installer for Linux. python3.6 for linux

2. In your Terminal window, run:

   bash Anaconda-latest-Linux-x86_64.sh

3. add path

   echo "# added by Anaconda3 installer
   export PATH="/home/never/anaconda3/bin:$PATH"" >> ~/.zshrc

Pytorch

conda install pytorch torchvision -c pytorch

tensorflow

1. Follow the instructions on the Anaconda download site to download and install Anaconda.

2. Create a conda environment named tensorflow to run a version of Python by invoking the following command:

   $ conda create -n tf python=3.6
   

3. Activate the conda environment by issuing the following command:

   $ source activate tf
    (tf)$  # Your prompt should change 
   

4. Issue a command of the following format to install TensorFlow inside your conda environment: (cuda9.0)

   sudo ln -s /usr/local/cuda/lib64/libcudnn.so.7.0.4 /usr/local/cuda/lib64/libcudnn.so.6pip install --ignore-installed --upgrade \ https://github.com/mind/wheels/releases/download/tf1.4-gpu-cuda9/tensorflow-1.4.0-cp36-cp36m-linux_x86_64.whl
   

5. MKL

   # If you don't have cmake
   sudo apt install cmake
   
   git clone https://github.com/01org/mkl-dnn.git
   cd mkl-dnn/scripts && ./prepare_mkl.sh && cd ..
   mkdir -p build && cd build && cmake .. && make
   sudo make install
   
   echo 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/lib' >> ~/.bashrc

6. Verification

   if import tensorflow —> error: *.so.8.0 not found

   for i in *so.9.1;do
   for> sudo ln -s $i ${i/%9.1/9.0}
   for> done

---

📚 Author : never

📅 Date : 2018.3.1

📧 Contact : zbh@ut-vision.org