how to reinstall cudnn？

Question

After install Cudnn.4.0.4, I find if I want to run the code I have downloaded, I need a higher version of Cudnn. Then I download cudnn-7.0-linux-x64-v4.0-prod.tgz and directly install by following order:

 $ sudo cp include/cudnn.h /usr/local/include
 $ sudo cp lib64/libcudnn.* /usr/local/lib
 $ sudo ln -sf /usr/local/lib/libcudnn.so.4.0.7 /usr/local/lib/libcudnn.so.4
 $ sudo ln -sf /usr/local/lib/libcudnn.so.4 /usr/local/lib/libcudnn.so
 $ sudo ldconfig -v

While I running sudo ldconfig -v, there is an error:

/sbin/ldconfig.real: /usr/local/cuda/lib64/libcudnn.so.4 is not a symbolic link

I run the source code again and have the same error before I reinstall cudnn:

 luajit: /home/zhfu/torch/install/share/lua/5.1/cudnn/ffi.lua:1287: These bindings are for version 4005 or above, while the loaded CuDNN is version: 4004  
 Are you using an older version of CuDNN?
 stack traceback:
    [C]: in function 'error'
    /home/zhfu/torch/install/share/lua/5.1/cudnn/ffi.lua:1287: in main chunk
    [C]: in function 'require'
    /home/zhfu/torch/install/share/lua/5.1/cudnn/init.lua:4: in main chunk
    [C]: in function 'require'
    ./main.lua:329: in main chunk
    [C]: at 0x00406670

Can somebody helps me to fix this problem?

Answer 1

Just delete all related files both in /usr/local/lib and in /usr/local/cuda/lib64 That's work for me.

Answer 2

In Linux, it looks like this:

sudo rm -r /usr/local/cuda-xxx/lib64/libcudnn*

cuda-xxx stand for cuda version: e.g 7.5 or 8.0

run the one of the following commands:

sudo rm -r /usr/local/cuda-7.5/lib64/libcudnn*

or

sudo rm -r /usr/local/cuda-8.0/lib64/libcudnn*