Memo

NVIDIA driver + CUDA + cuDNN インストール

Remove current nvidia driver and cuda

• dpkg -l | grep nvidia

• dpkg -l | grep cuda

• sudo apt-get --purge remove nvidia-*

• sudo apt-get --purge remove cuda-*

Install nvidia driver and cuda

• sudo add-apt-repository ppa:graphics-drivers/ppa

• sudo apt-get update

• sudo apt-get install nvidia-driver-XXX (You can check the recommended version using this command : ubuntu-drivers devices after installing sudo apt install ubuntu-drivers-common)

• Please go to the following web page: https://developer.nvidia.com/cuda-downloads and follow the Installation Instructions.

(this is an example for Ubuntu20.04, cuda11.0)

• wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin

• sudo mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600

• wget https://developer.download.nvidia.com/compute/cuda/11.0.3/local_installers/cuda-repo-ubuntu2004-11-0-local_11.0.3-450.51.06-1_amd64.deb

• sudo dpkg -i cuda-repo-ubuntu2004-11-0-local_11.0.3-450.51.06-1_amd64.deb

• sudo apt-key add /var/cuda-repo-ubuntu2004-11-0-local/7fa2af80.pub

• sudo apt-get update

• sudo apt-get -y install cuda

• export PATH="/usr/local/cuda/bin:$PATH"

• export LD_LIBRARY_PATH="/usr/local/cuda/lib64:$LD_LIBRARY_PATH"

• sudo reboot

Install cuDNN

• echo "deb https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu2004/x86_64 /" | sudo tee /etc/apt/sources.list.d/nvidia-ml.list

• sudo apt update

• sudo apt install libcudnn8-dev

• sudo reboot

or

• go to https://developer.download.nvidia.com/compute/machine-learning/repos/ and download an appropriate version.

• sudo dpkg -i libcudnn8_8.0.3.33-1+cuda11.0_amd64.deb

• sudo dpkg -i libcudnn8-dev_8.0.3.33-1+cuda11.0_amd64.deb

or

• go to https://developer.nvidia.com/rdp/cudnn-download and download an appropriate one.

• sudo dpkg -i cudnn-local-repo-${OS}-8.x.x.x_1.0-1_amd64.deb

• sudo apt-key add /var/cudnn-local-repo-*/7fa2af80.pub

• sudo apt-get update

• sudo apt-get install libcudnn8=8.x.x.x-1+cudaX.Y

• sudo apt-get install libcudnn8-dev=8.x.x.x-1+cudaX.Y

To check if the installation is successful, please run the following command:

• nvidia-smi

• nvcc -V

• cat /usr/include/cudnn.h | grep CUDNN_MAJOR - -A 2 or cat /usr/include/cudnn_version.h

• (reference : https://qiita.com/daichildren98/items/acc1abcca37dfa521fea)

or follow the section "Verifying The Install On Linux" in here.

2TB以上のSSDのパーティション作成・マウント

• parted /dev/sdX

• mklabel gpt

• mkpart

• Partition name? 1

• File system type? ext4

• Start? 0%

• End? 100%

• q

• sudo mkfs -t ext4 /dev/sdX1

• sudo mount /dev/sdX1 /mnt

Install Intel MKL to Ubuntu20.04

• pip uninstall numpy scipy

• sudo apt install intel-mkl

• create the following file at ~/numpy-site.cfg:

[mkl]

library_dirs = /usr/lib/x86_64-linux-gnu/

include_dirs = /usr/include/mkl

mkl_libs = mkl_rt

lapack_libs =

• create the following file at ~/.config/pip/pip.conf:

[install]

no-binary = numpy,scipy

• pip3 install --no-binary :all: numpy

• pip3 install --no-binary :all: scipy

Login loop (Ubuntu 14.04)

sudo apt-get install linux-image-4.4.0-93-generic

sudo apt-get install linux-headers-4.4.0-93

sudo apt-get install linux-headers-4.4.0-93-generic

sudo reboot

sudo apt-get remove nvidia* --purge

sudo apt-get autoremove

sudo apt-get update

(Download cuda driver like "NVIDIA-Linux-x86_64-384.111.run")

sudo service lightdm stop

sudo apt-get install linux-source

sudo bash ./NVIDIA-Linux-x86_64-384.111.run

sudo reboot now

古いカーネルの除去 (remove old kernels and headers)

uname -r で現状のバージョン確認

dpkg --get-selections | grep linux- でリスト表示

sudo apt-get autoremove --purge linux-headers-4.4.0-{128,130} で古いheader削除

sudo apt-get autoremove --purge linux-image-4.4.0-{128,130} で古いカーネル削除