- Notifications
You must be signed in to change notification settings - Fork23
HowTo | NVIDIA GPU and nvidia docker2
References:
If your Computer has NVIDIA Graphic cards, Install NVIDIA Graphic Driver to improve Graphics related performance.
Disablenouveau driver that is loaded by default as generic graphic driver.
Run following command as root or using sudo;
$ sudo update-pciids$ sudo lsmod| grep nouveaunouveau 1949696 0mxm_wmi 16384 1 nouveauwmi 32768 2 mxm_wmi,nouveauvideo 49152 1 nouveaui2c_algo_bit 16384 1 nouveauttm 106496 1 nouveaudrm_kms_helper 184320 1 nouveaudrm 491520 4 drm_kms_helper,ttm,nouveaulsmod: program to show the status of modules in the Linux Kernel
if the result containsnouveau, do following step
$ sudo vi /etc/modprobe.d/blacklist-nouveau.conf# add to the end (create new if it does not exist)blacklist nouveauoptions nouveau modeset=0$ sudo update-initramfs -u$ sudo reboot# confirm the cards on your computer$ sudo update-pciids$ lspci| grep VGA01:00.0 VGA compatible controller: NVIDIA Corporation GA102 [GeForce RTX 3090] (rev a1)
NVIDIA Driver is provided on Ubuntu official repository.Before installing, make sure the Driver Version for your Graphic Card on NVIDIA official site as follows:
For GeForce RTX 3090, driver version info from thehttps://www.nvidia.com/Download/driverResults.aspx/193095/en-us/
LINUX X64 (AMD64/EM64T) DISPLAY DRIVER
| Version: | 515.76 |
|---|---|
| Release Date: | 2022.9.20 |
| Operating System: | Linux 64-bit |
| Language: | English (US) |
| File Size: | 347.96 MB |
# install the driver for your card$ sudo apt -y install nvidia-driver-515# verify installation to show Graphic cards' status$ nvidia-smiIf you got the following error message, reboot your system or please refer to [Failed to initialize NVML: Driver/library version mismatch] link.
Failed to initialize NVML: Driver/library version mismatch
Refer to following reference for up-to-date information
#-----------------------------------# repo and GPG key configuration#-----------------------------------$ distribution=$(. /etc/os-release;echo$ID$VERSION_ID)\&& curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey| sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg\&& curl -s -L https://nvidia.github.io/libnvidia-container/$distribution/libnvidia-container.list|\ sed's#deb [https://#deb](https://github.com/#deb) [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] [https://#g'](https://github.com/#g') | \ sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list#-----------------------------------# nvidia-docker install with apt-get#-----------------------------------$ sudo apt-get update$ sudo apt-get install -y nvidia-docker2$ sudo apt-get install -y nvidia-container-toolkit#(Optional, Recommend to execute it for using GPU in Docker)#-----------------------------------# Configure the Docker daemon#-----------------------------------$ sudo nvidia-ctk runtime configure --runtime=docker#(Optional, Recommend to execute it for using GPU in Docker)#-----------------------------------# restart docker service#-----------------------------------$ sudo systemctl restart docker
$ docker run --rm --gpus all ubuntu:18.04 nvidia-smi
example output
+-----------------------------------------------------------------------------+| NVIDIA-SMI 460.73.01 Driver Version: 460.73.01 CUDA Version: 11.2 ||-------------------------------+----------------------+----------------------+| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC || Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. || | | MIG M. ||===============================+======================+======================|| 0 GeForce RTX 3090 Off | 00000000:01:00.0 On | N/A || 48% 44C P8 36W / 370W | 671MiB / 24267MiB | 4% Default || | | N/A |+-------------------------------+----------------------+----------------------+ +-----------------------------------------------------------------------------+| Processes: || GPU GI CI PID Type Process name GPU Memory || ID ID Usage ||=============================================================================|+-----------------------------------------------------------------------------+References:
You can specify GPU resource when executegit run commands as follows:
using --gpus option
$ docker run --rm --gpus all ubuntu:18.04 nvidia-smi
If you want to use specific GPU, you can use the GPU device number0,1,2, etc. instead ofall.
$ docker run --rm --gpus 0 ubuntu:18.04 nvidia-smi
using --runtime and environment variable
You should specifyNVIDIA_VISIBLE_DEVICES environment varialbe with--runtime option
$ docker run --rm --runtime=nvidia -e NVIDIA_VISIBLE_DEVICES=all ubuntu:18.04 nvidia-smi
Also you can specify the GPU device number instead ofall as follows:
$ docker run --rm --runtime=nvidia -e NVIDIA_VISIBLE_DEVICES=0 ubuntu:18.04 nvidia-smi
References
👉Note
- This page might contain some wide rendered images. If you want to see the whole contents with wider page, please useWide GitHub extension of Chrome.
- If you are experiencing the error on rendered images due to the low-bandwith Internet delay, please use F5 function key for refreshing current page.