E tensorflow/stream_executor/cuda/cuda_dnn.cc:352] Loaded runtime CuDNN library: 5005 (compatibility

WARNING:tensorflow:From /usr/local/lib/python2.7/dist-packages/tensorflow/python/util/tf_should_use.py:170: initialize_all_variables (from tensorflow.python.ops.variables) is deprecated and will be removed after 2017-03-02.
Instructions for updating:
Use `tf.global_variables_initializer` instead.
2018-01-02 17:14:41.226639: E tensorflow/stream_executor/cuda/cuda_dnn.cc:352] Loaded runtime CuDNN library: 5005 (compatibility version 5000) but source was compiled with 5110 (compatibility version 5100).  If using a binary install, upgrade your CuDNN library to match.  If building from sources, make sure the library loaded at runtime matches a compatible version specified during compile configuration.
2018-01-02 17:14:41.226836: F tensorflow/core/kernels/conv_ops.cc:671] Check failed: stream->parent()->GetConvolveAlgorithms(&algorithms) 

错误的主要原因是使用了cudnn5.0，应该使用cudnn5.1，链接地址：https://developer.nvidia.com/rdp/cudnn-download#a-collapseTwo

下面是在5.0 的基础上更新到5.1的过程：

 

删除之前安装的cudnn
$rm -rf /usr/local/cuda-8.0/targets/x86_64-linux/include/cudnn.h
$cd /usr/local/cuda/lib64
$sudo rm libcudnn.so
$sudo rm libcudnn.so.5
$sudo rm libcudnn.so.5.0.5

安装安装需要版本的cudnn，在终端cd到刚解压的cuda文件夹
$cd package
$sudo tar xvfcudnn-8.0-linux-x64-v5.1.tgz
$cd cuda/include
$sudo *.h /usr/local/cuda/include/
$cd ../lib64
$sudo cp libcudnn* /usr/local/cuda/lib64/
建立软链接（注意版本号换成你自己的）
$cd /usr/local/cuda/lib64
$sudo chmod +r libcudnn.so.5.1.10
$sudo ln -sf libcudnn.so.5.1.10 libcudnn.so.5
$sudo ln -sf libcudnn.so.5 libcudnn.so
$sudo ldconfig

接下来的错误可以看：http://blog.csdn.net/xhjj520/article/details/78857353