CODEWITHSUNDEEP
tensorflowgpucudnn
Peter Pan
Peter Pan

Reputation: 61

Loaded runtime CuDNN library: 8.5.0 but source was compiled with: 8.6.0

I have a problem with tensorflow. For weeks everything was working and now I get the error:

Loaded runtime CuDNN library: 8.5.0 but source was compiled with: 8.6.0. CuDNN library needs to have matching major version and equal or higher minor version. If using a binary install, upgrade your CuDNN library. If building from sources, make sure the library loaded at runtime is compatible with the version specified during compile configuration. 2023-07-27 15:54:24.546098: E tensorflow/compiler/xla/status_macros.cc:57] INTERNAL: RET_CHECK failure (tensorflow/compiler/xla/service/gpu/gpu_compiler.cc:618) dnn != nullptr

And I don't know how CuDNN library 8.5.0 is load. on PC only 8.9.1 is installed, I checked all cudnn_version.h and also reset path variable.

$LD_LIBRARY_PATH /mnt/c/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v11.8/lib/x64:/mnt/c/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v11.8/lib/x64::/home/user/miniconda3/envs/tf/lib/:/home/user/miniconda3/envs/tf/lib/python3. 9/site-packages/nvidia/cudnn/lib:/home/user/miniconda3/envs/tf/lib/:/home/user/miniconda3/envs/tf/lib/python3.9/site-packages/nvidia/cudnn/lib

I use in vs code with tensorflow the GPU over wsl2 which runs over ubuntu.

OrderedDict([('cpu_compiler', '/dt9/usr/bin/gcc'), ('cuda_compute_capabilities', ['sm_35', 'sm_50', 'sm_60', 'sm_70', 'sm_75', 'compute_80']), ('cuda_version', '11.8'), ('cudnn_version', '8'), ('is_cuda_build', True), ('is_rocm_build', False), ('is_tensorrt_build', True)])

Thanks for any help

I have checked the CuDNN version, updated the environment variable.

I hope to understand why cudnn 8.5.0 was loaded, and would like to know what I have to do to update the version to 8.6.0 or set it to the installed version 8.9.1.

Upvotes: 2

Views: 4335

Answers (1)

Peter Pan
Peter Pan

Reputation: 61

The problem was I had loaded pytorch to test something and when pip install pytorch the cudnn version was automatically changed with it.

The problem can be solved quite simply by re-running pip install nvidia-cudnn-cu11==8.6.0.163.

Upvotes: 4

Related Questions