NVIDIA CuDNN installation instructions

NVIDIA CuDNN installation instructions

CuDNN is a GPU computing Acceleration Solution designed specifically for the Deep Learning framework. Currently, the supported DL libraries include Caffe, ConvNet, and Torch7.

CuDNN can be obtained free of charge on the official website. You can download it after registering an account. The installation instructions are not found on the official website, and the downloaded package does not contain Readme. However, many instructions will be found on google. The basic principle is to add the lib file to the lib folder that the system can find, and add the header file to the include folder that the system can find. Add them to the CUDA folder (refer to here)

Tar-xzvf cudnn-6.5-linux-R1.tgzcd cp lib */usr/local/cuda/lib64/sudo cp cudnn. h/usr/local/cuda/include/

The database is still not found after execution, and an error is reported.

Error while loading shared libraries: libcudnn. so.6.5: cannot open shared object file: No such file or directory

The lib folder is in the system path, and ls-al is used to identify the File Permission problem. Therefore, use the following command to delete the soft connection first.

Cd/usr/local/cuda/lib64/sudo rm-rf libcudnn. so libcudnn. so.6.5

Then, modify the File Permission and create a new soft connection.

Sudo chmod u = rwx, g = rx, o = rx libcudnn. so.6.5.18 sudo ln-s libcudnn. so.6.5.18 libcudnn. so.6.5sudo ln-s libcudnn. so.6.5 libcudnn.

This article permanently updates the link address: