problem in Cudnn version change

AI & Data Science Deep Learning (Training & Inference) cuDNN
1

Hi,

Shortly, I am having a hard time about version change of cudnn in my TX2 board.(what i want to do is cudnn7.0.5 ->cudnn7.3.0)

I want to use GPU in the board to do deep learning with tensorflow.
But, when i use GPU for trainning i get “E tensorflow/stream_executor/cuda/cuda_dnn.cc:324] Loaded runtime CuDNN library: 7.0.5 but source was compiled with: 7.1.5” error message.
I had figured out what is the problem with me. but the previous solutions are not really helpful for me.

What i’ve done:

  1. download new version(7.3.0) of cudnn Tar file.

  2. “which nvcc” command say my cuda location is /usr/local/cuda-9.0/

  3. by following the documentation-2.3.1. i used these 3 commands([url]https://docs.nvidia.com/deeplearning/sdk/cudnn-install/index.html[/url])

    sudo cp cuda/include/cudnn.h /usr/local/cuda-9.0/include/
    sudo cp cuda/lib64/libcudnn* /usr/local/cuda-9.0/lib64
    sudo chmod a+r /usr/local/cuda-9.0/include/cudnn.h /usr/local/cuda-9.0/lib64/libcudnn*

result: “cat /usr/include/cudnn.h | grep CUDNN_MAJOR -A 2”, “cat /usr/local/cuda-9.0/include/cudnn.h | grep CUDNN_MAJOR -A 2” commands say that version is 7.3.0 but if i use “dpkg -l | grep cudnn”, the version is still 7.0.5

cf. CUDA: 9.0 , TX2-tegra-ubuntu, tensorflow-gpu: 1.9.0+nv18.8

2

hi,
I have the same issue.
When i use GPU i get " Loaded runtime CuDNN library: 7.3.1 but source was compiled with: 7.5.0".

3

hi,
I encountered the same problem. yesterday
I think the problem is why tensorflow-gpu is complied using other version of libcudnn than installed one on your jetson board.
Thus, I reinstalled tensorflow-gpu as specified in https://docs.nvidia.com/deeplearning/frameworks/install-tf-jetson-platform/index.html with the following command:
$ sudo pip3 install --pre --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v43 tensorflow-gpu

The problem has solved in my case.