TensorFlow：ImportError: libcudnn.so.*: cannot open shared object file: No such file or dictionary

环境：

• Ubuntu16.04
• Nvidia 384
• CUDA 8.0
• cuDNN 5

---

错误

TensorFlow、CUDA、cuDNN的版本关系小白我时常懵懵的，经常出现各种不支持的情况，比如今天安装了TensorFlow1.4，报错:ImportError: libcudnn.so.6: cannot open shared object file: No such file or dictionary

---

错误缘由

google了一圈发现， 问题出在 TensorFlow 1.4-gpu 是基于cuDNN6，需要的也就是libcudnn.so.6了。

---

解决方案：

• Just download cuDNN v6 and follow the steps (Tested on Ubuntu 16.04, CUDA toolkit 8.0 )https://developer.nvidia.com/cudnn
  $ tar xvzf cudnn-8.0-linux-x64-v5.1-ga.tgz
$ sudo cp -P cuda/include/cudnn.h /usr/local/cuda/include
$ sudo cp -P cuda/lib64/libcudnn* /usr/local/cuda/lib64
$ sudo chmod a+r /usr/local/cuda/include/cudnn.h /usr/local/cuda/lib64/libcudnn*
• Now set Path variables
• $ vim ~/.bashrc
翻到最底部加上：
export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64"
export CUDA_HOME=/usr/local/cuda

参考：http://blog.csdn.net/hungryof/article/details/52746279