Deep learning environment configuration - ubuntu20.04 installs cuda and cudnn

deep learning

1. Install CUDA

cuda is the programming language platform of NIVEA. If you want to use GPU, you must use cuda. ​​Download the installation file of cuda from here.
First select the appropriate version, and download the latest cuda 11.4 here.


Install according to the instructions above.
Because the NVIDIA driver is already installed, do not choose to install the NVIDIA driver here. The rest are default. As shown in the figure below, the first one is not selected.

After the installation is successful, you need to configure the environment variables. Otherwise, when using GPU acceleration, the GPU
configuration environment variables cannot be found:

gedit ~/.bashrc

Add the following statement at the end of the opened file and save:

export PATH=/usr/local/cuda-11.4/bin${
    
    PATH:+:${
    
    PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda-11.4/lib64${
    
    LD_LIBRARY_PATH:+:${
    
    LD_LIBRARY_PATH}}

Then update the environment variable

source ~/.bashrc

Enter nvcc -V to view related information

2. Install cudnn

Go to the official website to download the cudnn version matched with CUDA 11.4. To download cudnn, you need to register a NIVDIA account . Officials have given suggestions for matching cuda with cudnn. The sample download is cuDNN v8.8.2.

Select cuDNN Library for Linux as shown below, download cudnn-11.4-linux-x64-v8.2. 2.26.tgz and

decompress

tar -xvf cudnn-11.4-linux-x64-v8.2.2.26.tgz

Copy the relevant library files

sudo cp include/cudnn* /usr/local/cuda/include/ 
sudo cp 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*

View cudnn version

cat /usr/local/cuda/include/cudnn_version.h | grep CUDNN_MAJOR -A 2

This is the configuration is successful

monitor gpu status

watch -n 1 nvidia-smi