CuDNN library loaded at runtime: 5005 (compatibility version 5000), but the source was compiled with 5103 (compatibility version 5100)

Question

I have the following error. I use the conda setting of tensor flow. I'm struggling to use it with my GPU.

Loaded runtime CuDNN library: 5005 (compatibility version 5000) but source was compiled with 5103 (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. F tensorflow/core/kernels/conv_ops.cc:526] Check failed: stream->parent()->GetConvolveAlgorithms(&algorithms) Aborted (core dumped)

which returns nvcc /usr/local/cuda-7.5/bin/nvcc

Returns nvcc version of Cuda compilation tools, release 7.5, V7.5.17

I tried downloading CuDNN v5.1 and did the following, but it did not work either `` sudo cp lib * / usr / local / cuda-7.5 / lib64 / sudo cp include / cudnn.h / usr / local / cuda-7.5 / include / sudo ldconfig

`` ``

I also tried in another folder sudo cp lib* /usr/local/cuda/lib64/ sudo cp include/cudnn.h /usr/local/cuda/include/ sudo ldconfig

Answer 1

There is a good explanation of what this means - What makes the error: "CuDNN loaded runtime: 5005, but the source was compiled with 5103" means?

Short answer: you have CuDNN 5.0, but you must install CuDNN 5.1

Looks like what you are trying to do. It helped me just follow the instructions here - https://www.tensorflow.org/get_started/os_setup#optional_install_cuda_gpus_on_linux

Before doing this, I checked the contents of /usr/local/cuda/include/cudnn.h , and indeed, these lines were directed upwards, indicating that it was version 5.0.5

 #define CUDNN_MAJOR 5 #define CUDNN_MINOR 0 #define CUDNN_PATCHLEVEL 5 

If your /usr/local/cuda/include/cudnn.h already 5.1, then there is another CuDNN in another directory that is referenced. I have the following in my .bashrc - maybe try adding this or check out the Tensorflow instructions for adding.

 export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64" export CUDA_HOME="/usr/local/cuda"