import tensorflow report undefined symbol cudnnGetRNNParamsSize error

ubuntu x64, 14.04

import tensorflow report below error:

undefined symbol cudnnGetRNNParamsSize error

resolve:

1: ***      check your cuda version first  and cd to cuda installation  location , you can use:

--------------------------------------------------------------------------------

i:check cuda version:
cat /usr/local/cuda/version.txt

the install location should be :/usr/local/cuda-8.0

2: ***      check if libcudnn.so.6 as this symbol, use objdump

--------------------------------------------------------------------------------

objdump -TC  libcudnn.so.6, normally you can not find this symbol from this library.

3:***   under ubuntu , i install cuda-7.5 instead. you can do this by :

--------------------------------------------------------------------------------

sudo apt-get install cuda-7-5, this will uninstall cuda-8.0 first and install cuda 7.5 instead. for cuda 7.5 , you can find libcudnn.so under  /usr/lib/x86_64-linux-gnu/.   after this, you will find below error:

tensorflow：undefined symbol: cudnnSetRNNDescriptor_v6

now, please down cudnn-8.0-linux-x64-v6.0.tgz  here:
wget http://developer.download.nvidia.com/compute/redist/cudnn/v6.0/cudnn-8.0-linux-x64-v6.0.tgz

untar this archive to a folder for example named cudav6

copy and replace several files with flowing commands:

rm -f /usr/include/cudnn.h
rm -f /usr/lib/x86_64-linux-gnu/*libcudnn*
rm -f /usr/local/cuda-*/lib64/*libcudnn*

cp -P  /path/to/cudav6/include/cudnn.h /usr/include
cp -P /path/to/cudav6/lib64/libcudnn* /usr/lib/x86_64-linux-gnu/
chmod a+r /usr/lib/x86_64-linux-gnu/libcudnn*

after all operations,  you will find  import tensorflow work fine in python script.