INSTALL CENTOS STREAM 9 AND GPU AND CUDA

Recently, I need to do some research work about seismic wave equation solving and elastic wave field modeling. These work include some new method like PINN( Phisical neaural network) for partial differential operator. So, it is neccessary to have a CUDA installed on my Server. This server was bought in the year 2014.

Firstly, I downloaded the CENTOS STREAM 9 OS in its official website. And it is easy to clone one using dd command to realize a USB OR EFI USB system booting installler.

Secondly, after installing the OS, linux kernel and kernel header needs to be installed by using yum install command.  Some condition need to be satisfied. Such as, pip should be downloaded and could be used for the future using.The importance of the cycle explaining how to using yum repo and how to install kernel developing environment can be found on baidu.com or any other website like csdn and wechat.

Thirdly, following this website CentOS Stream 9 : NVIDIA Graphic Driver : Install : Server World, I can install the cuda driver for my gpu.

Some screenshot can be taken,using imagemagic,command: import /home/picture1.jpg

or import -window root /home/picture1.jpg

Fig1. cuda driver installed

From Fig1, it shows that nvidia gforce is off, so it needs the cmmand :nvidia-smi -pm 1 ,

to turn it on

File "/home/weiyiwen/anaconda3/lib/python3.12/site-packages/numba/cuda/cudadrv/devices.py", line 177, in _activate_context_for
    newctx = gpu.get_primary_context()
             ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/weiyiwen/anaconda3/lib/python3.12/site-packages/numba/cuda/cudadrv/driver.py", line 665, in get_primary_context
    met_requirement_for_device(self)
  File "/home/weiyiwen/anaconda3/lib/python3.12/site-packages/numba/cuda/cudadrv/driver.py", line 701, in met_requirement_for_device
    raise CudaSupportError("%s has compute capability < %s" %
numba.cuda.cudadrv.error.CudaSupportError: <CUDA device 0 'b'NVIDIA GeForce GT 640''> has comput