Skip to content

Latest commit

 

History

History
84 lines (83 loc) · 5.01 KB

install_cuda.md

File metadata and controls

84 lines (83 loc) · 5.01 KB
Raw

Install Nvidia CUDA-9.0 on Ubuntu 16.04 for Deep Learning

  • The latest version of CUDA is 9.1. However some deep learning frameworks are not yet ready for CUDA 9.1. The installation script of CUDA-9.1 is very similar to this one.
  • OS: Ubuntu 16.04 x86_64
  • (Optional) Uninstall old version CUDA Toolkit such as:
sudo apt-get purge cuda
sudo apt-get purge libcudnn6
sudo apt-get purge libcudnn6-dev
  • 刪除 cuda 可以參考下列指令 (移除 CUDA 版本 > 10.0)
sudo /usr/local/cuda-10.2/bin/cuda-uninstaller
  • 目前這樣的作法可以適用於 CUDA 10.1、CUDA 10.2。
  • 移除 CUDA 版本 <= 10.0 可以參考
sudo /usr/local/cuda-10.0/bin/uninstall_cuda_10.0.pl
  • Install CUDA Toolkit 9.0 and cuDNN 7.0 as follows:
wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/cuda-repo-ubuntu1604_9.0.176-1_amd64.deb
wget https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1604/x86_64/libcudnn7_7.0.5.15-1+cuda9.0_amd64.deb
wget https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1604/x86_64/libcudnn7-dev_7.0.5.15-1+cuda9.0_amd64.deb
wget https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1604/x86_64/libnccl2_2.1.4-1+cuda9.0_amd64.deb
wget https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1604/x86_64/libnccl-dev_2.1.4-1+cuda9.0_amd64.deb
sudo dpkg -i cuda-repo-ubuntu1604_9.0.176-1_amd64.deb
sudo dpkg -i libcudnn7_7.0.5.15-1+cuda9.0_amd64.deb
sudo dpkg -i libcudnn7-dev_7.0.5.15-1+cuda9.0_amd64.deb
sudo dpkg -i libnccl2_2.1.4-1+cuda9.0_amd64.deb
sudo dpkg -i libnccl-dev_2.1.4-1+cuda9.0_amd64.deb
sudo apt-get update
sudo apt-get install cuda=9.0.176-1
sudo apt-get install libcudnn7-dev
sudo apt-get install libnccl-dev
  • Reboot the system to load the NVIDIA drivers.
  • Set up the development environment by modifying the PATH and LD_LIBRARY_PATHvariables, also add them to the end of .bashrc file:
export PATH=/usr/local/cuda/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
  • 如果在 Ubuntu 14 想安裝 cuda 10 可參考 :
wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1404/x86_64/cuda-repo-ubuntu1404_10.0.130-1_amd64.deb
wget https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1404/x86_64/libcudnn7_7.4.1.5-1+cuda10.0_amd64.deb
wget https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1404/x86_64/libcudnn7-dev_7.4.1.5-1+cuda10.0_amd64.deb
wget https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1404/x86_64/libnccl2_2.3.5-2+cuda10.0_amd64.deb
wget https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1404/x86_64/libnccl-dev_2.3.5-2+cuda10.0_amd64.deb
sudo dpkg -i cuda-repo-ubuntu1404_10.0.130-1_amd64.deb
sudo dpkg -i libcudnn7_7.4.1.5-1+cuda10.0_amd64.deb
sudo dpkg -i libcudnn7-dev_7.4.1.5-1+cuda10.0_amd64.deb
sudo dpkg -i libnccl2_2.3.5-2+cuda10.0_amd64.deb
sudo dpkg -i libnccl-dev_2.3.5-2+cuda10.0_amd64.deb
sudo apt-get update
sudo apt-get install cuda=10.0.130-1
sudo apt-get install libcudnn7-dev
sudo apt-get install libnccl-dev
  • 如果在 Ubuntu 18.04 想安裝 cuda 10.2 可參考 :
wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/cuda-10-2_10.2.89-1_amd64.deb
wget https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64/libcudnn7_7.6.5.32-1+cuda10.2_amd64.deb
wget https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64/libcudnn7-dev_7.6.5.32-1+cuda10.2_amd64.deb
wget https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64/libnccl2_2.7.8-1+cuda10.2_amd64.deb
wget https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64/libnccl-dev_2.7.8-1+cuda10.2_amd64.deb
sudo dpkg -i cuda-10-2_10.2.89-1_amd64.deb
sudo dpkg -i libcudnn7_7.6.5.32-1+cuda10.2_amd64.deb
sudo dpkg -i libcudnn7-dev_7.6.5.32-1+cuda10.2_amd64.deb
sudo dpkg -i libnccl2_2.7.8-1+cuda10.2_amd64.deb
sudo dpkg -i libnccl-dev_2.7.8-1+cuda10.2_amd64.deb
sudo apt-get update
sudo apt-get install cuda=10.2.89-1
sudo apt-get install libcudnn7-dev
sudo apt-get install libnccl-dev
  • 各 lib 要求的版本沒有硬性要求，只要結尾都是 cuda10.0_amd64.deb

參考資料 :

  1. https://yangcha.github.io/CUDA90/
  2. nvidia-driver 安裝 / 更新 : https://gist.github.com/wangruohui/df039f0dc434d6486f5d4d098aa52d07#install-nvidia-graphics-driver-via-apt-get
  3. 這個有安裝成功，強烈建議使用 .run 安裝，不要用 apt-get 裝 nvidia-driver : https://www.itread01.com/content/1549538125.html
  4. 可用指令 lsmod 確認 nvidia-driver 相關程序是否被開啟 : https://blog.csdn.net/u011668104/article/details/79560381
  5. How to disable Nouveau kernel driver : https://askubuntu.com/questions/841876/how-to-disable-nouveau-kernel-driver
  6. [Linux] 簡單的移除 CUDA 心得筆記 https://clay-atlas.com/blog/2020/04/09/%E7%B0%A1%E5%96%AE%E7%9A%84%E7%A7%BB%E9%99%A4-cuda-%E5%BF%83%E5%BE%97%E7%AD%86%E8%A8%98/