Automatic setup of nvidia docker and some convenience features on Ubuntu 16.04

configure_ubuntu.sh

#!/bin/sh
APPLICATIONS="$HOME/Apps"
mkdir -p $APPLICATIONS
LOCALBIN="$HOME/.local/bin"
mkdir -p $LOCALBIN

install_fun() {
	prog=$1
	if ! command -v $1 > /dev/null; then
		echo "$1 is not available. Trying to install it."
		sudo apt-get install $1
	fi
}

install_appimage() {
	url=$1
	name=$2
	wget $url -O $APPLICATIONS/$name.appimage
	chmod +x $APPLICATIONS/$name.appimage
	ln -s $APPLICATIONS/$name.appimage $LOCALBIN/$name

}

# install fish from latest ppa
sudo apt-add-repository -y ppa:fish-shell/release-2
sudo add-apt-repository -y ppa:aacebedo/fasd
sudo apt-get update && sudo apt-get install -y fish fasd

# install aws tools
sudo apt-get install -y awscli

# install latest neovim as appimage
sudo apt-get install -y python3-pip
sudo pip3 install neovim
install_appimage https://github.com/neovim/neovim/releases/download/v0.3.1/nvim.appimage nvim

# Install Configuration files
install_fun git
install_fun stow
if ! [ -d "Configurations" ]; then
	git clone https://github.com/xiamaz/Configurations.git
fi

# Create configs
cd ~/Configurations
make fish
make tmux
make neovim
make git

install_nvidia_cuda_ml.sh

#!/bin/sh

NVIDIA_REPO=/etc/apt/sources.list.d/nvidia-ml.list
install_nvidia_ml_repo () {
	if [ -e $NVIDIA_REPO ]; then
		echo "Repo already added."
		return
	fi
	sudo sh -c "echo \"deb http://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1604/x86_64 /\" > $NVIDIA_REPO"
}

remove_nvidia_ml_repo () {
	rm -r $NVIDIA_REPO
}


CUDA_VERSION=9.0
CUDNN_VERSION="7.3.1.20-1+cuda$CUDA_VERSION"
NCCL_VERSION="2.3.5-2+cuda$CUDA_VERSION"
install_nccl() {
	nccl_file=nccl_${NCCL_VERSION}_x86_64.txz
	if ! [ -e $nccl_file ]; then
		echo "Please download $nccl_file"
		return
	fi
	sudo cp $nccl_file /usr/local
	sudo tar xvf /usr/local/$nccl_file -C /usr/local/
	sudo mv /usr/local/$(basename $nccl_file .txz) /usr/local/nccl-2.3
	sudo rm /usr/local/$nccl_file
}
# Install cuda, cudnn and nccl, eg everything we need to make tensorflow run
install_cuda() {
	if [ -d /usr/local/cuda-$CUDA_VERSION ]; then
		echo "Cuda $CUDA_VERSION already installed"
	else
		sudo apt-get update && sudo apt-get install cuda-$CUDA_VERSION
	fi

	sudo apt-get update
	sudo apt-get install libcudnn7=$CUDNN_VERSION libcudnn7-dev=$CUDNN_VERSION
	# libnccl2=$NCCL_VERSION libnccl-dev=$NCCL_VERSION
	install_nccl
}

install_nvidia_ml_repo

install_cuda

install_nvidia_docker.sh

#!/bin/sh
set -e  # exit script when one command fails
set -u  # unset variables will cause failure

install_dockerce(){
	sudo apt-get install -y apt-transport-https ca-certificates curl software-properties-common
	curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo apt-key add -
	sudo add-apt-repository -y "deb [arch=amd64] https://download.docker.com/linux/ubuntu \
		$(lsb_release -cs) stable"
	sudo apt-get update && sudo apt-get install -y docker-ce
}

install_nvidia(){
	install_dockerce
	curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
	distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
	curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | \
	  sudo tee /etc/apt/sources.list.d/nvidia-docker.list
	sudo apt-get update

	sudo apt-get install -y nvidia-docker2
	sudo pkill -SIGHUP dockerd
}

nvtest() {
	docker run --runtime=nvidia --rm nvidia/cuda:9.0-base nvidia-smi
}

# Test nvidia-smi with the latest official CUDA image
case $1 in
	install)
		install_nvidia
		;;
	test)
		nvtest
		;;
	*)
		echo "Unknown argument $1. Valid options are: install, test."
		;;
esac

setup_python36_tensorflow.sh

#!/bin/sh

TENSORFLOW_ENV=tensorflow_p36
CONDADIR=$HOME/miniconda

# CONDA Installation functions

# install conda on the current user
install_conda() {
	if [ -d $CONDADIR ]; then
		echo "Conda already installed in $CONDADIR"
		return
	fi
	wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh -O miniconda.sh
	bash ~/miniconda.sh -b -p $CONDADIR
	# create tf environment
	$CONDADIR/conda create -y -n $TENSORFLOW_ENV python=3.6
}

CONDA_BASHCONF=$HOME/.bashrc
conda_bash() {
	if grep "$CONDADIR" $CONDA_BASHCONF; then
		echo "Conda already in bashrc."
		return
	fi
	echo "export PATH=\"$CONDADIR/bin:\$PATH\"" >> $CONDA_BASHCONF
}

# configure fish shell for conda use
CONDA_FISHCONF=$HOME/.config/fish/conf.d/conda.fish
conda_fish() {
	if [ -e $CONDA_FISHCONF ]; then
		echo "Conda fish config already exists in $CONDA_FISHCONF"
		return
	fi
	echo "source $CONDADIR/etc/fish/conf.d/conda.fish\nconda activate $TENSORFLOW_ENV" > $CONDA_FISHCONF
}

# Tensorflow Installation functions

BAZELVER=0.17.2

install_bazel() {
	if command -v bazel > /dev/null; then
		echo "Bazel already installed."
		return
	fi
	if ! [ -e bazel-installer.sh ]; then
		wget https://github.com/bazelbuild/bazel/releases/download/$BAZELVER/bazel-$BAZELVER-installer-linux-x86_64.sh -O bazel-installer.sh
	fi
	# install bazel to user
	bash bazel-installer.sh --user
}

TENSORFLOW_VERSION="r1.11"

build_tensorflow() {
	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

	if ! [ -d tensorflow ]; then
		git clone https://github.com/tensorflow/tensorflow.git
	fi
	cd tensorflow
	git checkout $TENSORFLOW_VERSION && git pull
	# Configure if not configured yet
	if ! [ -e .tf_configure.bazelrc ]; then
		./configure
	fi
	# bazel build --config=opt --config=cuda //tensorflow/tools/pip_package:build_pip_package
	./bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/tensorflow_pkg
	pip install /tmp/tensorflow_pkg/*.whl
}


# install_conda
# conda_bash
# conda_fish

# install_bazel
build_tensorflow

Updated to include more scripts containg almost all steps required to build a native tensorflow package.