目錄
- 本篇概覽
- 特別問(wèn)題說(shuō)明
- 準(zhǔn)備工作
- 下載和安裝Nvidia驅(qū)動(dòng)
- 安裝CUDA
- 安裝cuDNN
本篇概覽
自己有一臺(tái)2015年的聯(lián)想筆記本,顯卡是GTX950M,已安裝ubuntu 16.04 LTS桌面版,為了使用其GPU完成deeplearning4j的訓(xùn)練工作,自己動(dòng)手安裝了CUDA和cuDNN,在此將整個(gè)過(guò)程記錄下來(lái),以備將來(lái)參考,整個(gè)安裝過(guò)程分為以下幾步:
- 準(zhǔn)備工作
- 安裝Nvidia驅(qū)動(dòng)
- 安裝CUDA
- 安裝cuDNN
特別問(wèn)題說(shuō)明
- 按照一般步驟,在安裝完Nvidia顯卡驅(qū)動(dòng)后,會(huì)提示對(duì)應(yīng)的CUDA版本,接下來(lái)按照提示的版本安裝CUDA,例如我這里提示的是11.2,正常情況下,我應(yīng)該安裝11.2版本的CUDA
- 但是我選擇9.1版本就行安裝,因?yàn)橹暗拈_(kāi)發(fā)中發(fā)現(xiàn)deeplearning4j使用了11.2的SDK后,啟動(dòng)應(yīng)用會(huì)有ClassNotFound的錯(cuò)誤,此問(wèn)題至今未修復(fù)(慚愧,欣宸水平如此之低…),因此,我在Nvidia驅(qū)動(dòng)提示11.2版本的情況下,依然安裝了9.1版本,后來(lái)在此環(huán)境運(yùn)行deeplearning4j應(yīng)用一切正常
- 如果您沒(méi)有我這類問(wèn)題,完全可以按照驅(qū)動(dòng)指定的版本來(lái)安裝CUDA,具體的操作步驟稍后會(huì)詳細(xì)說(shuō)到;
準(zhǔn)備工作
- 接下來(lái)的操作,除了在網(wǎng)頁(yè)下載,其余都是ssh遠(yuǎn)程連接到ubuntu機(jī)器操作的,ssh登錄的帳號(hào)為普通帳號(hào),并非root
- 如果已有驅(qū)動(dòng),請(qǐng)先刪除
sudo apt-get remove --purge nvidia*
禁用nouveau驅(qū)動(dòng)(很重要),用vi打開(kāi)文件/etc/modprobe.d/blacklist.conf,在尾部增加以下內(nèi)容,然后保存退出:
blacklist nouveau
blacklist lbm-nouveau
options nouveau modeset=0
alias nouveau off
alias lbm-nouveau off
關(guān)閉nouveau:
echo options nouveau modeset=0 | sudo tee -a /etc/modprobe.d/nouveau-kms.conf
更新initramfs:
執(zhí)行reboot重啟電腦
重啟后,執(zhí)行以下命令,應(yīng)該不會(huì)有任何輸出,證明nouveau已經(jīng)禁用:
獲取Kernel source:
sudo apt-get install linux-source
安裝過(guò)程中顯示信息如下圖:

根據(jù)上圖紅框中的信息,可知內(nèi)核版本號(hào)為,于是執(zhí)行以下命令:
sudo apt-get install linux-headers-4.4.0-210-generic
下載和安裝Nvidia驅(qū)動(dòng)
訪問(wèn)Nvidia網(wǎng)站,地址https://www.nvidia.cn/Download/index.aspx?lang=cn,然后選擇對(duì)應(yīng)的顯卡和操作系統(tǒng),我的選擇如下圖所示:

點(diǎn)擊上圖搜索按鈕后,進(jìn)入下圖頁(yè)面,點(diǎn)擊下載:

下載得到的文件名為NVIDIA-Linux-x86_64-460.84.run
關(guān)閉圖形頁(yè)面:
sudo service lightdm stop
給驅(qū)動(dòng)文件增加可執(zhí)行權(quán)限:
sudo chmod a+x NVIDIA-Linux-x86_64-460.84.run
開(kāi)始安裝:
sudo ./NVIDIA-Linux-x86_64-460.84.run -no-x-check -no-nouveau-check -no-opengl-files
遇到下圖,選擇紅框:

遇到下圖,直接回車:

恢復(fù)圖形頁(yè)面:
sudo service lightdm start
執(zhí)行命令nvidia-smi,如果驅(qū)動(dòng)安裝成功,會(huì)顯示以下內(nèi)容:
will@lenovo:~/temp/202106/20$ nvidia-smi
Sun Jun 20 09:02:11 2021
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 460.84 Driver Version: 460.84 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 GTX 950M Off | 00000000:01:00.0 Off | N/A |
| N/A 41C P0 N/A / N/A | 0MiB / 4046MiB | 1% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| No running processes found |
+-----------------------------------------------------------------------------+
從上述內(nèi)容可見(jiàn)CUDA Version: 11.2表示該驅(qū)動(dòng)對(duì)應(yīng)的CUDA版本應(yīng)該是11.2,正如前面所說(shuō),我這邊遇到了問(wèn)題,因此接下來(lái)會(huì)安裝9.1版本,但是您可以選擇安裝11.2
安裝CUDA
瀏覽器訪問(wèn)https://developer.nvidia.com/cuda-toolkit-archive,點(diǎn)擊紅框中的鏈接:

如下圖,下載Linux版本:

繼續(xù)選擇x86_64:

選擇具體的Linux版本及其版本號(hào):

要下載的東西不少,一個(gè)安裝程序和三個(gè)補(bǔ)丁:

上述四個(gè)文件的下載地址整理如下:
https://developer.download.nvidia.cn/compute/cuda/9.1/secure/Prod/local_installers/cuda_9.1.85_387.26_linux.run?P0Ntu_6NLtuuEMm6fJRk1W5vl4KM7oaT1oFW870zKJ-zDw2ckKntFLOE6klRJfw2CmTa8z3Q390_6urlgc6LqjoqlIFW9gvfvDCusnINYplLaw1u8lRY8R4oVNtpNzaXU4BQcHjvdb6c6rjq20dktCcRd4640woXt1yHmD95v1Du7wdBBXq2eOY
https://developer.download.nvidia.cn/compute/cuda/9.1/secure/Prod/patches/1/cuda_9.1.85.1_linux.run?yeXf_7wIGlHAUw--E_YVLQZRgXv0x2i043woJVY-ydXU5Kyhc-eYQf5JmL-4mvYmlvPYCEc5RhT2sDWscX20CJbdOwpkt30kWb9vx8E4oIlajDQ3MVPvXdiKKsIOBUx-h0q0N0jSkNn80VMhW-nk8jwvRY_e6MuFzqWBaPk
https://developer.download.nvidia.cn/compute/cuda/9.1/secure/Prod/patches/2/cuda_9.1.85.2_linux.run?5jGZxNigaOJkaaPbMagjhSW7ebQvYGyYoqe2vBxZ1eV8qp2BzXJLxIPgAo11UgWhORirQkdJGq5b8eFh4aShBVUTmuPaasvRiMCKDZw5yjjIobGQrCEyU-LFO59AbrRER57Mxa0T1Sc97fC80IOZq8Ox2repjn7A3oYVgd8
https://developer.download.nvidia.cn/compute/cuda/9.1/secure/Prod/patches/3/cuda_9.1.85.3_linux.run?CxWimJTC-XROYihig-UZmH62odbJInf1fmxTZ_bsW1nQ0Zz5cL5r8qLmlMR_1j2rVhk3j8Z5lS6dpArt8frjGHH2MeVn5TefMoclam8udm-RSMMmqHXYE66hHN2D0drVEdtCwe8ZrEIYb2rpucaz9svCFE8Z319mge4Ju94
下載完畢后,執(zhí)行命令chmod a+x *.run為上述四個(gè)文件增加可執(zhí)行權(quán)限
安裝CUDA:
sudo sh cuda_9.1.85_387.26_linux.run
遇到license時(shí),像是用vi工具那樣,輸入":",再輸入"q"回車,就能跳過(guò)license閱讀,執(zhí)行真正的安裝操作了:

接下來(lái)是一系列提問(wèn),每一個(gè)提問(wèn)的回答如下圖,千萬(wàn)注意紅框中的問(wèn)題一定要選擇n:

安裝完成后輸出以下內(nèi)容:
Installing the CUDA Toolkit in /usr/local/cuda-9.1 ...
Missing recommended library: libGLU.so
Missing recommended library: libX11.so
Missing recommended library: libXi.so
Missing recommended library: libXmu.so
Missing recommended library: libGL.so
Installing the CUDA Samples in /home/will ...
Copying samples to /home/will/NVIDIA_CUDA-9.1_Samples now...
Finished copying samples.
===========
= Summary =
===========
Driver: Not Selected
Toolkit: Installed in /usr/local/cuda-9.1
Samples: Installed in /home/will, but missing recommended libraries
Please make sure that
- PATH includes /usr/local/cuda-9.1/bin
- LD_LIBRARY_PATH includes /usr/local/cuda-9.1/lib64, or, add /usr/local/cuda-9.1/lib64 to /etc/ld.so.conf and run ldconfig as root
To uninstall the CUDA Toolkit, run the uninstall script in /usr/local/cuda-9.1/bin
Please see CUDA_Installation_Guide_Linux.pdf in /usr/local/cuda-9.1/doc/pdf for detailed information on setting up CUDA.
***WARNING: Incomplete installation! This installation did not install the CUDA Driver. A driver of version at least 384.00 is required for CUDA 9.1 functionality to work.
To install the driver using this installer, run the following command, replacing CudaInstaller> with the name of this run file:
sudo CudaInstaller>.run -silent -driver
Logfile is /tmp/cuda_install_13425.log
打開(kāi)文件~/.bashrc,在尾部增加以下兩行(LD_LIBRARY_PATH如果已經(jīng)存在,請(qǐng)參考PATH的寫法改成追加):
export PATH=/usr/local/cuda-9.1/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda-9.1/lib64
執(zhí)行命令source ~/.bashrc使配置生效
執(zhí)行命令su -切換到root帳號(hào),執(zhí)行以下命令(不要用sudo,而是切到root帳號(hào)):
sudo echo "/usr/local/cuda-9.1/lib64" >> /etc/ld.so.conf
再以root身份執(zhí)行以下命令:
執(zhí)行命令exit退出root身份,現(xiàn)在又是普通帳號(hào)的身份了
執(zhí)行命令nvcc -V檢查CUDA版本,注意參數(shù)V是大寫:
will@lenovo:~$ nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2017 NVIDIA Corporation
Built on Fri_Nov__3_21:07:56_CDT_2017
Cuda compilation tools, release 9.1, V9.1.85
安裝第一個(gè)補(bǔ)丁:
sudo sh cuda_9.1.85.1_linux.run
安裝第二個(gè)補(bǔ)丁:
sudo sh cuda_9.1.85_387.26_linux.run
安裝第三個(gè)補(bǔ)丁:
sudo sh cuda_9.1.85_387.26_linux.run
安裝cuDNN
瀏覽器訪問(wèn)https://developer.nvidia.com/zh-cn/cudnn:

按提示登錄,如果沒(méi)有帳號(hào)請(qǐng)注冊(cè)一個(gè),登錄后進(jìn)入下載頁(yè)面,需要點(diǎn)擊下圖紅框位置才有能見(jiàn)到老版本:

選擇與CUDA匹配的版本:

下載后解壓,得到文件夾cuda,然后執(zhí)行以下命令:
sudo cp cuda/include/cudnn.h /usr/local/cuda/include/
sudo cp cuda/lib64/libcudnn* /usr/local/cuda/lib64/
sudo chmod a+r /usr/local/cuda/include/cudnn.h
sudo chmod a+r /usr/local/cuda/lib64/libcudnn*
執(zhí)行檢查確認(rèn)的命令cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2,如果安裝順利會(huì)有以下輸出:
#define CUDNN_MAJOR 7
#define CUDNN_MINOR 1
#define CUDNN_PATCHLEVEL 3
--
#define CUDNN_VERSION (CUDNN_MAJOR * 1000 + CUDNN_MINOR * 100 + CUDNN_PATCHLEVEL)
#include "driver_types.h"
至此,Ubuntu16安裝CUDA(9.1)和cuDNN已經(jīng)完成了,希望能給您一些參考。
到此這篇關(guān)于Ubuntu16安裝CUDA(9.1)和cuDNN的實(shí)現(xiàn)步驟(圖文)的文章就介紹到這了,更多相關(guān)Ubuntu16安裝CUDA(9.1)和cuDNN內(nèi)容請(qǐng)搜索腳本之家以前的文章或繼續(xù)瀏覽下面的相關(guān)文章希望大家以后多多支持腳本之家!
您可能感興趣的文章:- 詳解Ubuntu16.04安裝nvidia驅(qū)動(dòng)+CUDA+cuDNN的教程