origin
- nvidia-smi is inconsistent with nvcc --version version,
please refer to https://blog.csdn.net/weixin_39518984/article/details/111406728 - The experiment requires a higher version of cuda
Tensorflow cuda cudnn version corresponding reference https://blog.csdn.net/weixin_45184581/article/details/124583991
uninstall old cuda
sudo /usr/local/cuda-9.0/bin/uninstall_cuda_9.0.pl
Go to the /usr/local/cuda-9.0 folder to check, generally speaking, there are some cudnn files, delete it
sudo rm -rf /usr/local/cuda-9.0
install new cuda
Download address https://developer.nvidia.com/cuda-toolkit-archive
Download and install according to the command prompts respectively
When installing:
Do not install the driver
Modify environment variables
sudo gedit ~/.bashrc
Change the original content to
export PATH="/usr/local/cuda-11.2/bin:$PATH"
export LD_LIBRARY_PATH="/usr/local/cuda-11.2/lib64:$LD_LIBRARY_PATH"
Save and exit execution
source ~/.bashrc
verify cuda
cd /usr/local/cuda-11.6
chmod 777 samples
cd samples
git clone https://gitee.com/liwuhaoooo/cuda-samples.git
#或者git clone samples里面给的那个地址
after downloading
cd /usr/local/cuda/samples/1_Utilities/deviceQuery
sudo make
./deviceQuery
Successful installation
install cudnn
The download address https://developer.nvidia.com/zh-cn/cudnn
must be registered and logged in first.
After downloading, unzip and enter the unzipped folder
sudo cp include/cudnn.h /usr/local/cuda-11.6/include
sudo cp lib/libcudnn* /usr/local/cuda-11.6/lib64
sudo chmod a+r /usr/local/cuda-11.6/include/cudnn.h
sudo chmod a+r /usr/local/cuda-11.6/lib64/libcudnn*