TensorFlow + jupyter notebook + Nvidia DIY setup __tensorflow

TensorFlow + jupyter notebook + Nvidia DIY setup

This is a detailed installation step on Ubuntu 17.04 that requires sudo permissions. Nvidia driver Settings

Install NVIDIA Drive Warehouse

sudo add-apt-repository ppa:graphics-drivers/ppa
sudo apt-get update
sudo apt-get upgrade
sudo apt-get Install build-essential cmake g++ gfortran git pkg-config python-dev software-properties-common

Install Cuda 8.0 from NVIDIA, including installer and patch 2.

sudo dpkg-i cuda-repo-ubuntu1604-8-0-local-*amd64.deb
sudo apt-get update
sudo apt-get install Cuda

Setting up a CUDA environment for local Users

Echo ' Export path=/usr/local/cuda/bin: $PATH ' >> ~/.BASHRC
echo ' Export
ld_library_path=/usr/local/ CUDA/LIB64: $LD _library_path ' >>
~/.bashrc
Source ~/.BASHRC

Test Cuda

Nvcc-v

Finally install CUDNN, this need to note and Cuda version corresponds to

CD ~/downloads/
tar xvf cudnn*.tgz
cd cuda
sudo cp */*.h/usr/local/cuda/include/
sudo cp */libcudnn*/ usr/local/cuda/lib64/
sudo chmod a+r/usr/local/cuda/lib64/libcudnn*

Anaconda Settings

Install Anaconda, select Python3.6 version

CD ~/downloads/
bash anaconda3-4.4.0-linux-x86_64.sh

The anaconda is usually installed in the/OPT directory, and if it is installed in the home directory, then the environment directory needs to be determined after activating the environment.

SOURCE Activate Tf-gpu
echo $PATH

TensorFlow Settings

When we install the Anaconda, we need to set up Jupyter and install the TensorFlow GPU.

Conda Create--name Tf-gpu python=3.6
source activate Tf-gpu
pip install--ignore-installed
--upgrade HTTPS://STORAGE.GOOGLEAPIS.COM/TENSORFLOW/LINUX/GPU/TENSORFL
OW_GPU-1.3.0-CP36-CP36M-LINUX_X86_64.WHL

jupyter Settings

After setting the Tendorflow, we install the jupyter and some of the packages that will be used

Conda Install Jupyter notebook NumPy Pandas