Install GPU/CUDA/

카테고리 없음

Install GPU/CUDA/

Sumin Lim 2019. 1. 7. 12:24

1. Checking Server spec

suminlim@universe ~$ lsb_release -a

No LSB modules are available.

Distributor ID: Ubuntu

Description: Ubuntu 16.04.5 LTS

Release: 16.04

Codename: xenial

suminlim@universe ~$ python --version

Python 2.7.12

suminlim@universe ~$ cat /var/log/dpkg.log.1 | grep nvidia

2018-12-10 17:08:13 install libnvidia-container1:amd64 <none> 1.0.0-1

2018-12-10 17:08:13 status half-installed libnvidia-container1:amd64 1.0.0-1

2018-12-10 17:08:13 status unpacked libnvidia-container1:amd64 1.0.0-1

2018-12-10 17:08:13 install libnvidia-container-tools:amd64 <none> 1.0.0-1

2018-12-10 17:08:13 status half-installed libnvidia-container-tools:amd64 1.0.0-1

2018-12-10 17:08:13 status unpacked libnvidia-container-tools:amd64 1.0.0-1

2018-12-10 17:08:14 install nvidia-container-runtime-hook:amd64 <none> 1.4.0-1

2018-12-10 17:08:14 status half-installed nvidia-container-runtime-hook:amd64 1.4.0-1

2018-12-10 17:08:14 status unpacked nvidia-container-runtime-hook:amd64 1.4.0-1

2018-12-10 17:08:14 install nvidia-container-runtime:amd64 <none> 2.0.0+docker18.09.0-1

2018-12-10 17:08:14 status half-installed nvidia-container-runtime:amd64 2.0.0+docker18.09.0-1

2018-12-10 17:08:14 status unpacked nvidia-container-runtime:amd64 2.0.0+docker18.09.0-1

2018-12-10 17:08:14 install nvidia-docker2:all <none> 2.0.3+docker18.09.0-1

2018-12-10 17:08:14 status half-installed nvidia-docker2:all 2.0.3+docker18.09.0-1

2018-12-10 17:08:14 status unpacked nvidia-docker2:all 2.0.3+docker18.09.0-1

2018-12-10 17:08:14 configure libnvidia-container1:amd64 1.0.0-1 <none>

2018-12-10 17:08:14 status unpacked libnvidia-container1:amd64 1.0.0-1

2018-12-10 17:08:14 status half-configured libnvidia-container1:amd64 1.0.0-1

2018-12-10 17:08:14 status installed libnvidia-container1:amd64 1.0.0-1

2018-12-10 17:08:14 configure libnvidia-container-tools:amd64 1.0.0-1 <none>

2018-12-10 17:08:14 status unpacked libnvidia-container-tools:amd64 1.0.0-1

2018-12-10 17:08:14 status half-configured libnvidia-container-tools:amd64 1.0.0-1

2018-12-10 17:08:14 status installed libnvidia-container-tools:amd64 1.0.0-1

2018-12-10 17:08:15 configure nvidia-container-runtime-hook:amd64 1.4.0-1 <none>

2018-12-10 17:08:15 status unpacked nvidia-container-runtime-hook:amd64 1.4.0-1

2018-12-10 17:08:15 status half-configured nvidia-container-runtime-hook:amd64 1.4.0-1

2018-12-10 17:08:15 status installed nvidia-container-runtime-hook:amd64 1.4.0-1

2018-12-10 17:08:15 configure nvidia-container-runtime:amd64 2.0.0+docker18.09.0-1 <none>

2018-12-10 17:08:15 status unpacked nvidia-container-runtime:amd64 2.0.0+docker18.09.0-1

2018-12-10 17:08:15 status half-configured nvidia-container-runtime:amd64 2.0.0+docker18.09.0-1

2018-12-10 17:08:15 status installed nvidia-container-runtime:amd64 2.0.0+docker18.09.0-1

2018-12-10 17:08:15 configure nvidia-docker2:all 2.0.3+docker18.09.0-1 <none>

2018-12-10 17:08:15 status unpacked nvidia-docker2:all 2.0.3+docker18.09.0-1

2018-12-10 17:08:15 status half-configured nvidia-docker2:all 2.0.3+docker18.09.0-1

2018-12-10 17:08:15 status installed nvidia-docker2:all 2.0.3+docker18.09.0-1

2. Open MPI

https://www.open-mpi.org/faq/?category=building#easy-build

shell$ gunzip -c openmpi-4.0.0.tar.gz | tar xf -
shell$ cd openmpi-4.0.0
shell$ ./configure --prefix=/usr/local
<...lots of output...>
shell$ make all install

Conventional Installation

https://towardsdatascience.com/tensorflow-gpu-installation-made-easy-use-conda-instead-of-pip-52e5249374bc

pytorch와 tensorflow-gpu를 설치한다

conda create --name tf_gpu tensorflow_gpu

conda activate tf_gpu

conda install tensorflow-gpu

conda install -y pytorch torchvision -c pytorch

source deactivate

https://dreamgonfly.github.io/2018/01/30/conda-pytorch.html

3. CUDA install

Installation Instructions:

`sudo dpkg -i cuda-repo-ubuntu1604-10-0-local-10.0.130-410.48_1.0-1_amd64.deb`
`sudo apt-key add /var/cuda-repo-<version>/7fa2af80.pub`
`sudo apt-get update`
`sudo apt-get install cuda`

PyTorch

Conda base

https://dreamgonfly.github.io/2018/01/30/conda-pytorch.html

https://beerensahu.wordpress.com/2018/03/20/pytorch-installation-guide-for-ubuntu/

$ python3 -m venv pytorch-env

$ source pytorch-env/bin/activate

Reference links for Installation

http://www.kwangsiklee.com/2017/07/우분투-16-04에서-cuda-성공적으로-설치하기/

http://walking-on-the-life.tistory.com/9

저작자표시 (새창열림)