Could NOT find CUDNN (missing: CUDNN_LIBRARY_PATH CUDNN_INCLUDE_PATH)

[nevergone123]

Wonderful works!!!
I was trying Step 2.3: cmake . -B build
However，I encountered the following problem:

(chatsim) root@admin:~/AImodel/wenke/ChatSim/chatsim/background/mcnerf# cmake . -B build
-- Obtained CUDA architectures automatically from installed GPUs
-- Targeting CUDA architectures: 80
-- Module support is disabled.
-- Version: 9.1.1
-- Build type: Release
-- Caffe2: CUDA detected: 11.7
-- Caffe2: CUDA nvcc is: /usr/local/cuda-11.7/bin/nvcc
-- Caffe2: CUDA toolkit directory: /usr/local/cuda-11.7
-- Caffe2: Header version is: 11.7
-- Could NOT find CUDNN (missing: CUDNN_LIBRARY_PATH CUDNN_INCLUDE_PATH)
CMake Warning at External/libtorch/share/cmake/Caffe2/public/cuda.cmake:120 (message):
Caffe2: Cannot find cuDNN library. Turning the option off
Call Stack (most recent call first):
External/libtorch/share/cmake/Caffe2/Caffe2Config.cmake:92 (include)
External/libtorch/share/cmake/Torch/TorchConfig.cmake:68 (find_package)
CMakeLists.txt:39 (find_package)

-- /usr/local/cuda-11.7/lib64/libnvrtc.so shorthash is d833c4f3
-- Autodetected CUDA architecture(s): 8.0 8.0 8.0 8.0 8.0 8.0 8.0 8.0
-- Added CUDA NVCC flags for: -gencode;arch=compute_80,code=sm_80
CMake Error at External/libtorch/share/cmake/Caffe2/Caffe2Config.cmake:100 (message):
Your installed Caffe2 version uses cuDNN but I cannot find the cuDNN
libraries. Please set the proper cuDNN prefixes and / or install cuDNN.
Call Stack (most recent call first):
External/libtorch/share/cmake/Torch/TorchConfig.cmake:68 (find_package)
CMakeLists.txt:39 (find_package)

-- Configuring incomplete, errors occurred!
See also "/root/AImodel/wenke/ChatSim/chatsim/background/mcnerf/build/CMakeFiles/CMakeOutput.log".
See also "/root/AImodel/wenke/ChatSim/chatsim/background/mcnerf/build/CMakeFiles/CMakeError.log".

Do you have any idea? Many thanks!!

[yifanlu0227]

You should install CUDNN as the error message says.

option 1

Go to CUDNN download page to download and install it.

option 2

You can also download CUDNN via conda, but remember to set CUDNN_LIBRARY_PATH and CUDNN_INCLUDE_PATH to your conda environment like

export CUDNN_INCLUDE_PATH=$CONDA_PREFIX/include:$CUDNN_INCLUDE_PATH 
export CUDNN_LIBRARY_PATH=$CONDA_PREFIX/lib:$CUDNN_LIBRARY_PATH

[nevergone123]

@yifanlu0227 , cool! Option 1 solved the problem. Thanks again！