How To Setup NVIDIA CUDA Drvier on Ubuntu
- 3. 02. Install Ubuntu 16.04 ● It’ll Based on Ubuntu 16.04 64Bit – All Environments are Containerized Using Docker (also nvidiadocker will help using dedicated Graphics Processor Unit) – For Example, You can use Dockerized Tensorflow image for Train & Serve ● & There are Several types of tensorflow images ● https://hub.docker.com/r/tensorflow/tensorflow/tags/ ● Install using USB Flash Driver
- 5. 03. Install NVIDIA Driver ● Unfortuneately, NVIDIA Drivers are SHIT ● First, You should download CUDA Driver on Link BELOW ● https://developer.nvidia.com/cudadownloads ● Select distribution and click “Download” Button ● (I prefer download local deb archive to increase download time)
- 7. 03. Install NVIDIA Driver ● (All Commands are based on “Base Installer”) ● First, Go to downloaded directory (using cd, ls, etc) ● Second, Using dpkg to decompress deb archive – `sudo dpkg i cudarepoubuntu160491local_9.1.851_amd64.deb` ● Third, Watch Terminal and Copy and paste String Like this – `sudo aptkey add /var/cudarepo<version>/7fa2af80.pub` – DO NOT COPY & PASTE On this Slide, You Should Copy and Paste On TERMINAL
- 8. 03. Install NVIDIA Driver ● Fourth, Update APT using – `sudo aptget update` ● Fifth, Install CUDA using APT – `sudo aptget install cuda` – It’ll Remove Old NVIDIA and Install Newer Driver ● Lastly, Reboot Computer – After Reboot, Type `nvidiasmi` on Terminal to check graphics card’s Driver are Successfully Installed