Skip to content

Instantly share code, notes, and snippets.

@matheustguimaraes

matheustguimaraes/install-cuda-cudnn.md

Last active March 20, 2023 05:08
Show Gist options
  • Star 75 You must be signed in to star a gist
  • Fork 20 You must be signed in to fork a gist
  • Save matheustguimaraes/43e0b65aa534db4df2918f835b9b361d to your computer and use it in GitHub Desktop.
Save matheustguimaraes/43e0b65aa534db4df2918f835b9b361d to your computer and use it in GitHub Desktop.
Download ZIP
Install CUDA 10.0 and cuDNN v7.4.2 on Ubuntu 16.04
Raw
install-cuda-cudnn.md

Install CUDA 10.0 and cuDNN v7.4.2 on Ubuntu 16.04

Originally posted on: https://matheustguimaraes.com/blog/cuda-cudnn-ubuntu-installation

Install the latest NVIDIA driver

Update package lists, download and install NVIDIA driver

sudo apt-get update
sudo add-apt-repository ppa:graphics-drivers/ppa
sudo apt install nvidia-410

Reboot

Restart the computer

reboot

Testing

Lets test if all worked well

nvidia-smi

If appears something like:

Sun Jan 27 15:33:47 2019       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 410.78       Driver Version: 410.78       CUDA Version: 10.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce MX130       Off  | 00000000:01:00.0 Off |                  N/A |
| N/A   49C    P5    N/A /  N/A |    495MiB /  2004MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      1091      G   /usr/lib/xorg/Xorg                           285MiB |
|    0      2014      G   compiz                                       103MiB |
|    0      2263      G   ...quest-channel-token=6295865577498169014   103MiB |
+-----------------------------------------------------------------------------+

The driver was installed.

Download CUDA Toolkit

Download the CUDA Toolkit on NVIDIA official website https://developer.nvidia.com/cuda-downloads

You have to make some choices about your machine to download the file

My machine:

  • Linux
  • x86_64
  • Ubuntu
  • 16.04
  • runfile (local)

Go to the download folder

After download the file, go to the download folder:

cd Downloads

Run the file

sudo sh cuda_10.0.130_410.48_linux.run

You have to make some choices in the terminal:

  • Install NVIDIA Accelerated Graphics Driver for Linux-x86_64 384.81? n
  • Install the CUDA 10.0 Toolkit? y
  • Do you want to install a symbolic link at /usr/local/cuda? y
  • Install the CUDA 10.0 Samples? y

Leave the rest as default. ENTER

Testing

Let's test if everything worked well

cd /usr/local/cuda/samples
sudo make -k
./deviceQuery

If appears something like:

./deviceQuery Starting...

 CUDA Device Query (Runtime API) version (CUDART static linking)

Detected 1 CUDA Capable device(s)

Device 0: "GeForce MX130"
  CUDA Driver Version / Runtime Version          10.0 / 10.0
  CUDA Capability Major/Minor version number:    5.0
  Total amount of global memory:                 2004 MBytes (2101870592 bytes)
  ( 3) Multiprocessors, (128) CUDA Cores/MP:     384 CUDA Cores
  GPU Max Clock rate:                            1189 MHz (1.19 GHz)
  Memory Clock rate:                             2505 Mhz
  Memory Bus Width:                              64-bit
  L2 Cache Size:                                 1048576 bytes
  Maximum Texture Dimension Size (x,y,z)         1D=(65536), 2D=(65536, 65536), 3D=(4096, 4096, 4096)
  Maximum Layered 1D Texture Size, (num) layers  1D=(16384), 2048 layers
  Maximum Layered 2D Texture Size, (num) layers  2D=(16384, 16384), 2048 layers
  Total amount of constant memory:               65536 bytes
  Total amount of shared memory per block:       49152 bytes
  Total number of registers available per block: 65536
  Warp size:                                     32
  Maximum number of threads per multiprocessor:  2048
  Maximum number of threads per block:           1024
  Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
  Max dimension size of a grid size    (x,y,z): (2147483647, 65535, 65535)
  Maximum memory pitch:                          2147483647 bytes
  Texture alignment:                             512 bytes
  Concurrent copy and kernel execution:          Yes with 1 copy engine(s)
  Run time limit on kernels:                     Yes
  Integrated GPU sharing Host Memory:            No
  Support host page-locked memory mapping:       Yes
  Alignment requirement for Surfaces:            Yes
  Device has ECC support:                        Disabled
  Device supports Unified Addressing (UVA):      Yes
  Device supports Compute Preemption:            No
  Supports Cooperative Kernel Launch:            No
  Supports MultiDevice Co-op Kernel Launch:      No
  Device PCI Domain ID / Bus ID / location ID:   0 / 1 / 0
  Compute Mode:
     < Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 10.0, CUDA Runtime Version = 10.0, NumDevs = 1
Result = PASS

The library was installed!

Install cuDNN

Download cuDNN v7.4.2 (Dec 14, 2018), for CUDA 10.0

Download the cuDNN on NVIDIA official website https://developer.nvidia.com/cudnn

NOTE: You have to sign up at NVIDIA website before

Download the three packages:

  • cuDNN Runtime Library for Ubuntu16.04 (Deb)
  • cuDNN Developer Library for Ubuntu16.04 (Deb)
  • cuDNN Code Samples and User Guide for Ubuntu16.04 (Deb)

Go to downloads folder

cd Downloads

Install the .deb packages

sudo dpkg -i libcudnn7_7.4.2.24-1+cuda10.0_amd64.deb
sudo dpkg -i libcudnn7-dev_7.4.2.24-1+cuda10.0_amd64.deb 
sudo dpkg -i libcudnn7-doc_7.4.2.24-1+cuda10.0_amd64.deb

Export CUDA path

export LD_LIBRARY_PATH=/usr/local/cuda-10.0/lib64:$LD_LIBRARY_PATH
export PATH=/usr/local/cuda-10.0/bin:$PATH

Reload bash

source ~/.bashrc

Testing

Go in your /usr/src and copy the 'cudnn_samples_v7' to any folder you want, to test if all worked well. In my case, I paste it in 'Desktop' folder.

cd Desktop/cudnn_samples_v7/mnistCUDNN/
make clean && make
./mnistCUDNN

If appears something like:

cudnnGetVersion() : 7402 , CUDNN_VERSION from cudnn.h : 7402 (7.4.2)
Host compiler version : GCC 5.4.0
There are 1 CUDA capable devices on your machine :
device 0 : sms  3  Capabilities 5.0, SmClock 1189.0 Mhz, MemSize (Mb) 2004, MemClock 2505.0 Mhz, Ecc=0, boardGroupID=0
Using device 0

Testing single precision
Loading image data/one_28x28.pgm
Performing forward propagation ...
Testing cudnnGetConvolutionForwardAlgorithm ...
Fastest algorithm is Algo 1
Testing cudnnFindConvolutionForwardAlgorithm ...
^^^^ CUDNN_STATUS_SUCCESS for Algo 0: 0.034528 time requiring 0 memory
^^^^ CUDNN_STATUS_SUCCESS for Algo 1: 0.038624 time requiring 3464 memory
^^^^ CUDNN_STATUS_SUCCESS for Algo 2: 0.040992 time requiring 57600 memory
^^^^ CUDNN_STATUS_SUCCESS for Algo 4: 0.174656 time requiring 207360 memory
^^^^ CUDNN_STATUS_SUCCESS for Algo 7: 0.440832 time requiring 2057744 memory
Resulting weights from Softmax:
0.0000000 0.9999399 0.0000000 0.0000000 0.0000561 0.0000000 0.0000012 0.0000017 0.0000010 0.0000000 
Loading image data/three_28x28.pgm
Performing forward propagation ...
Resulting weights from Softmax:
0.0000000 0.0000000 0.0000000 0.9999288 0.0000000 0.0000711 0.0000000 0.0000000 0.0000000 0.0000000 
Loading image data/five_28x28.pgm
Performing forward propagation ...
Resulting weights from Softmax:
0.0000000 0.0000008 0.0000000 0.0000002 0.0000000 0.9999820 0.0000154 0.0000000 0.0000012 0.0000006 

Result of classification: 1 3 5

Test passed!

Testing half precision (math in single precision)
Loading image data/one_28x28.pgm
Performing forward propagation ...
Testing cudnnGetConvolutionForwardAlgorithm ...
Fastest algorithm is Algo 1
Testing cudnnFindConvolutionForwardAlgorithm ...
^^^^ CUDNN_STATUS_SUCCESS for Algo 0: 0.036960 time requiring 0 memory
^^^^ CUDNN_STATUS_SUCCESS for Algo 1: 0.051584 time requiring 3464 memory
^^^^ CUDNN_STATUS_SUCCESS for Algo 2: 0.066176 time requiring 28800 memory
^^^^ CUDNN_STATUS_SUCCESS for Algo 4: 0.199936 time requiring 207360 memory
^^^^ CUDNN_STATUS_SUCCESS for Algo 7: 0.516160 time requiring 2057744 memory
Resulting weights from Softmax:
0.0000001 1.0000000 0.0000001 0.0000000 0.0000563 0.0000001 0.0000012 0.0000017 0.0000010 0.0000001 
Loading image data/three_28x28.pgm
Performing forward propagation ...
Resulting weights from Softmax:
0.0000000 0.0000000 0.0000000 1.0000000 0.0000000 0.0000714 0.0000000 0.0000000 0.0000000 0.0000000 
Loading image data/five_28x28.pgm
Performing forward propagation ...
Resulting weights from Softmax:
0.0000000 0.0000008 0.0000000 0.0000002 0.0000000 1.0000000 0.0000154 0.0000000 0.0000012 0.0000006 

Result of classification: 1 3 5

Test passed!

You installed it everything with great success.

Everything is ready for you to use the GPU and do great things.

Have a nice day!

@MuhammadAsadJaved
Copy link

How we can check cudnn version if installed using .deb? if we installed using .run we can check using
cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2

@Harsha-1729
Copy link

Hi
Thank you. It worked for me.

After Installing I could see 2 versions of cudnn as follows

image

image

When I have installed tensorflow-gpu==1.5 version and try to check the version, ERROR: 'Failed to load the native TensorFlow runtime' is shown. So when I have checked the reason in https://www.tensorflow.org/install/source#common_installation_problems
I could see this information:

image

So, Now I need only 1 cudnn to be on my pc. I want to uninstall cudnn 7.4.2.
How can I do that

thanks in advance
regards
harsha

@Harsha-1729
Copy link

Hi
Thank you. It worked for me.

After Installing I could see 2 versions of cudnn as follows

image

image

When I have installed tensorflow-gpu==1.5 version and try to check the version, ERROR: 'Failed to load the native TensorFlow runtime' is shown. So when I have checked the reason in https://www.tensorflow.org/install/source#common_installation_problems
I could see this information:

image

So, Now I need only 1 cudnn to be on my pc. I want to uninstall cudnn 7.4.2.
How can I do that

thanks in advance
regards
harsha

uninstall command: sudo apt-get --purge remove libcudnn7

@adijindal30
Copy link

Thanks bro!!

@nandini211995
Copy link

Thank you so much :)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment