MACHINE-SETUP.md

Summary

This is a tutorial to guide you to setup an ubuntu machine. You are welcome!

• Refer: https://gist.github.com/matheustguimaraes/43e0b65aa534db4df2918f835b9b361d

11 Steps

Step1 Install NVIDIA DRIVER

sudo apt-get upgrade sudo apt-get update sudo add-apt-repository ppa:graphics-drivers/ppa sudo apt install nvidia-410

Step2 Reboot

reboot

Step3 Check NVIDIA DRIVER

nvidia-smi

Step4 Install CUDA 10.0

• Download Linux/x86_64/Ubuntu/16.04/runfile(local): https://developer.nvidia.com/cuda-10.0-download-archive?target_os=Linux&target_arch=x86_64&target_distro=Ubuntu&target_version=1604&target_type=runfilelocal

sudo sh cuda_10.0.130_410.48_linux.run

    Install NVIDIA Accelerated Graphics Driver for Linux-x86_64 384.81? n
    Install the CUDA 10.0 Toolkit? y
    Do you want to install a symbolic link at /usr/local/cuda? y
    Install the CUDA 10.0 Samples? y

Step5 Test CUDA

cd /usr/local/cuda/samples
sudo make -k
cd /usr/local/cuda/samples/1_Utilities/deviceQuery
./deviceQuery

Step6 Install cuDNN v7.4.2 (Dec 14, 2018), for CUDA 10.0

Download the following:

• cuDNN Runtime Library for Ubuntu16.04 (Deb)
• cuDNN Developer Library for Ubuntu16.04 (Deb)
• cuDNN Code Samples and User Guide for Ubuntu16.04 (Deb)
• sudo dpkg -i libcudnn7_7.4.2.24-1+cuda10.0_amd64.deb
• sudo dpkg -i libcudnn7-dev_7.4.2.24-1+cuda10.0_amd64.deb
• sudo dpkg -i libcudnn7-doc_7.4.2.24-1+cuda10.0_amd64.deb

The add those to ~/.bashrc

export LD_LIBRARY_PATH=/usr/local/cuda-10.0/lib64:$LD_LIBRARY_PATH
export PATH=/usr/local/cuda-10.0/bin:$PATH
source ~/.bashrc

Step7 Test cuDNN

cd Desktop/cudnn_samples_v7/mnistCUDNN/
make clean && make
./mnistCUDNN

Step8 Install Anaconda Anaconda3-2020.02-Linux-x86_64.sh

• Follow https://docs.anaconda.com/anaconda/install/linux/

source ~/.bashrc
conda config --set auto_activate_base False

Step9 Create an python 3.7 environment

conda create -n py3 python==3.7

Step10 Install tensorflow-gpu 1.14

conda activate py3
pip install tensorflow-gpu==1.14

Step11 Test GPU availability

python
import tensorflow as tf
tf.test.is_gpu_available(
    cuda_only=False, min_cuda_compute_capability=None
)

~> You shall see True