Trying to compile dlib 19.20 with cuda 11 and cudnn 8, or cuda 10.1 and cudnn 7.6.4

[spiderdab]

Environment:
Pop-Os (Ubuntu 20.04)
Gcc 9 and 8 (tried both)
Cmake 3.16.3
dlib 19.20.99
python 3.8.2
NVIDIA Quadro P600

Expected Behavior

Compiling with cuda..

Current Behavior

.....
-- Found cuDNN: /usr/lib/x86_64-linux-gnu/libcudnn.so
-- Building a CUDA test project to see if your compiler is compatible with CUDA...
-- Checking if you have the right version of cuDNN installed.
-- *** Found cuDNN, but it looks like the wrong version so dlib will not use it. ***
-- *** Dlib requires cuDNN V5.0 OR GREATER. Since cuDNN is not found DLIB WILL NOT USE CUDA.
.....

• Where did you get dlib:
  git clone https://github.com/davisking/dlib.git

I first tried downloading the latest (and suggested) version of cuda from Nvidia site, which was Cuda 11. then downloaded cuDNN for Cuda 11, which was 8.0.0 (runtime and dev .deb packets) installed them following Nvidia method.

when compiling "cudnn_samples_v8" everything works, so I think installations went ok.
but no way to get dlib compiled with Cuda.

I've tried to uninstall cuda 11 and cudnn 8 and install cuda 10.1 and cudnn 7.6 (suggested for cuda 10.1) but the result is the same.
Every time I erased the build directory, and compiled in 2 ways:
as per dlib instructions, using:
$ sudo python3 setup.py install
or:
$ cmake .. -DDLIB_USE_CUDA=1 -DUSE_AVX_INSTRUCTIONS=1

Any suggestion?
Thanks.

[davisking]

That test for cudnn is done by building the dlib/cmake_utils/test_for_cudnn cmake project and checking if the build succeeds. So if you build dlib/cmake_utils/test_for_cudnn specifically you can see why it failed. I will note that CUDA 11 is still a release candidate, and Ubuntu 20.04 is not an officially supported distribution by NVIDIA, so that might be why it's not building. Most of the times I try to get CUDA working on a non-supported distribution it's an exercise in frustration.

…

On Tue, Jun 9, 2020 at 3:28 PM spiderdab ***@***.***> wrote: Environment: Pop-Os (Ubuntu 20.04) Gcc 9 and 8 (tried both) Cmake 3.16.3 dlib 19.20.99 python 3.8.2 NVIDIA Quadro P600 Expected Behavior Compiling with cuda.. Current Behavior ..... -- Found cuDNN: /usr/lib/x86_64-linux-gnu/libcudnn.so -- Building a CUDA test project to see if your compiler is compatible with CUDA... -- Checking if you have the right version of cuDNN installed. -- *** Found cuDNN, but it looks like the wrong version so dlib will not use it. *** -- *** Dlib requires cuDNN V5.0 OR GREATER. Since cuDNN is not found DLIB WILL NOT USE CUDA. ..... - *Where did you get dlib*: git clone https://github.com/davisking/dlib.git I first tried downloading the latest (and suggested) version of cuda from Nvidia site, which was Cuda 11. then downloaded cuDNN for Cuda 11, which was 8.0.0 (runtime and dev .deb packets) installed them following Nvidia method. when compiling "cudnn_samples_v8" everything works, so I think installations went ok. but no way to get dlib compiled with Cuda. I've tried to uninstall cuda 11 and cudnn 8 and install cuda 10.1 and cudnn 7.6 (suggested for cuda 10.1) but the result is the same. Every time I erased the build directory, and compiled in 2 ways: as per dlib instructions, using: $ sudo python3 setup.py install or: $ cmake .. -DDLIB_USE_CUDA=1 -DUSE_AVX_INSTRUCTIONS=1 Any suggestion? Thanks. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#2100>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABPYFR2QMY577NZ6TQEESC3RV2EOLANCNFSM4NZWP7FA> .

[spiderdab]

Thanks for your time.
Trying to compile "test_for_cudnn" I have this result:

dab@pop-os:/home/dab/dlib/dlib/cmake_utils/test_for_cudnn/build$ cmake ..
-- The C compiler identification is GNU 8.4.0
-- The CXX compiler identification is GNU 8.4.0
-- Check for working C compiler: /usr/bin/cc
-- Check for working C compiler: /usr/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Check for working CXX compiler: /usr/bin/c++
-- Check for working CXX compiler: /usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Looking for pthread.h
-- Looking for pthread.h - found
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Failed
-- Looking for pthread_create in pthreads
-- Looking for pthread_create in pthreads - not found
-- Looking for pthread_create in pthread
-- Looking for pthread_create in pthread - found
-- Found Threads: TRUE
-- Found CUDA: /usr/local/cuda (found suitable version "10.1", minimum required is "7.5")
-- Looking for cuDNN install...
-- Found cuDNN: /usr/lib/x86_64-linux-gnu/libcudnn.so
-- C++11 activated.
-- Configuring done
-- Generating done
-- Build files have been written to: /home/dab/dlib/dlib/cmake_utils/test_for_cudnn/build
dab@pop-os:/home/dab/dlib/dlib/cmake_utils/test_for_cudnn/build$ ls
CMakeCache.txt CMakeFiles cmake_install.cmake Makefile
dab@pop-os:~/dlib/dlib/cmake_utils/test_for_cudnn/build$ make
Scanning dependencies of target cudnn_test
[ 50%] Building CXX object CMakeFiles/cudnn_test.dir/home/dab/dlib/dlib/cuda/cudnn_dlibapi.cpp.o
In file included from /home/dab/dlib/dlib/cuda/cudnn_dlibapi.cpp:10:
/usr/local/include/cudnn.h:59:10: fatal error: cudnn_version.h: File o directory non esistente
#include "cudnn_version.h"
^~~~~~~~~~~~~~~~~
compilation terminated.
make[2]: *** [CMakeFiles/cudnn_test.dir/build.make:63: CMakeFiles/cudnn_test.dir/home/dab/dlib/dlib/cuda/cudnn_dlibapi.cpp.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:76: CMakeFiles/cudnn_test.dir/all] Error 2
make: *** [Makefile:84: all] Error 2

should I find "cudnn_version.h" and copy in ".../dlib/cuda/" folder?

[davisking]

No, don't copy any cudnn files into the dlib folder Installing cudnn should put that header file in a standard place the compiler can find it. Something is wrong with your install if gcc can't find that header.

[spiderdab]

I've found that header here:
dab@pop-os:~$ sudo locate cudnn_version.h [sudo] password di dab: /usr/include/cudnn_version.h

it looks like a common folder.. Can you point me on how to add that path to cmake, or what else I can do?

thanks.

[spiderdab]

I could finally install dlib with cuda and cudnn into Pop!_OS 20.04 (ubuntu) using cuda 10.1 and dlib 7.6.4
into the Nvidia download page you can download 3 .deb packets but they are not enough, since as written above the correct headers are not found. so I also downloaded one archive which name is simply "cuDNN Library for Linux" and copied 2 folders (include and libx86) from this archive to cuda folder using these 2 commands:
`dab@pop-os:~/cudnn-10.1-linux-x64-v7.6.4.38/cuda$ sudo cp -r lib64/ /usr/local/cuda-10.1/targets/x86_64-linux/lib/

dab@pop-os:~/cudnn-10.1-linux-x64-v7.6.4.38/cuda$ sudo cp -r include/* /usr/local/cuda-10.1/targets/x86_64-linux/include/
`
after that DLIB compilation went all as expected.

Thanks, Davide.

[davisking]

Huh, you should definitely not need to do this stuff. But like I was saying, nvidia doens't support cuda on that version of ubuntu yet, and maybe cmake isn't updated to try and make it work either. Inside dlib/CMakeLists.txt you can find the statements that find cuda. You could add to the paths there and maybe that would help though.

[spiderdab]

I've tried that before, but didn't work. (for sure my fault..)
Thank you very much for your time and help.

[alexmarkley]

I'm also having a similar (but not identical issue), apparently related to some API changes in cuDNN.

alex@ubsidian:~/Build/dlib/dlib_git/dlib/cmake_utils/test_for_cudnn/build$ cmake ..
-- The C compiler identification is GNU 9.3.0
-- The CXX compiler identification is GNU 9.3.0
-- Check for working C compiler: /usr/bin/cc
-- Check for working C compiler: /usr/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Check for working CXX compiler: /usr/bin/c++
-- Check for working CXX compiler: /usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Looking for pthread.h
-- Looking for pthread.h - found
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Failed
-- Looking for pthread_create in pthreads
-- Looking for pthread_create in pthreads - not found
-- Looking for pthread_create in pthread
-- Looking for pthread_create in pthread - found
-- Found Threads: TRUE  
-- Found CUDA: /usr/local/cuda (found suitable version "11.0", minimum required is "7.5") 
-- Looking for cuDNN install...
-- Found cuDNN: /usr/lib/x86_64-linux-gnu/libcudnn.so
-- C++11 activated.
-- Configuring done
-- Generating done
-- Build files have been written to: /home/alex/Build/dlib/dlib_git/dlib/cmake_utils/test_for_cudnn/build
alex@ubsidian:~/Build/dlib/dlib_git/dlib/cmake_utils/test_for_cudnn/build$ make
Scanning dependencies of target cudnn_test
[ 50%] Building CXX object CMakeFiles/cudnn_test.dir/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp.o
/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp: In member function ‘void dlib::cuda::tensor_conv::setup(const dlib::tensor&, const dlib::tensor&, int, int, int, int)’:
/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp:850:57: error: ‘CUDNN_CONVOLUTION_FWD_PREFER_FASTEST’ was not declared in this scope; did you mean ‘CUDNN_CONVOLUTION_BWD_FILTER_ALGO_3’?
  850 |                         dnn_prefer_fastest_algorithms()?CUDNN_CONVOLUTION_FWD_PREFER_FASTEST:CUDNN_CONVOLUTION_FWD_NO_WORKSPACE,
      |                                                         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp:43:33: note: in definition of macro ‘CHECK_CUDNN’
   43 |     const cudnnStatus_t error = call;                                         \
      |                                 ^~~~
/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp:850:94: error: ‘CUDNN_CONVOLUTION_FWD_NO_WORKSPACE’ was not declared in this scope; did you mean ‘CUDNN_CONVOLUTION_FWD_ALGO_WINOGRAD’?
  850 |                         dnn_prefer_fastest_algorithms()?CUDNN_CONVOLUTION_FWD_PREFER_FASTEST:CUDNN_CONVOLUTION_FWD_NO_WORKSPACE,
      |                                                                                              ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp:43:33: note: in definition of macro ‘CHECK_CUDNN’
   43 |     const cudnnStatus_t error = call;                                         \
      |                                 ^~~~
/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp:844:29: error: ‘cudnnGetConvolutionForwardAlgorithm’ was not declared in this scope; did you mean ‘cudnnGetConvolutionForwardAlgorithm_v7’?
  844 |                 CHECK_CUDNN(cudnnGetConvolutionForwardAlgorithm(
      |                             ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp:43:33: note: in definition of macro ‘CHECK_CUDNN’
   43 |     const cudnnStatus_t error = call;                                         \
      |                                 ^~~~
/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp:872:57: error: ‘CUDNN_CONVOLUTION_BWD_DATA_PREFER_FASTEST’ was not declared in this scope; did you mean ‘CUDNN_CONVOLUTION_BWD_DATA_ALGO_FFT’?
  872 |                         dnn_prefer_fastest_algorithms()?CUDNN_CONVOLUTION_BWD_DATA_PREFER_FASTEST:CUDNN_CONVOLUTION_BWD_DATA_NO_WORKSPACE,
      |                                                         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp:43:33: note: in definition of macro ‘CHECK_CUDNN’
   43 |     const cudnnStatus_t error = call;                                         \
      |                                 ^~~~
/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp:872:99: error: ‘CUDNN_CONVOLUTION_BWD_DATA_NO_WORKSPACE’ was not declared in this scope; did you mean ‘CUDNN_CONVOLUTION_BWD_DATA_ALGO_WINOGRAD’?
  872 |                         dnn_prefer_fastest_algorithms()?CUDNN_CONVOLUTION_BWD_DATA_PREFER_FASTEST:CUDNN_CONVOLUTION_BWD_DATA_NO_WORKSPACE,
      |                                                                                                   ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp:43:33: note: in definition of macro ‘CHECK_CUDNN’
   43 |     const cudnnStatus_t error = call;                                         \
      |                                 ^~~~
/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp:866:29: error: ‘cudnnGetConvolutionBackwardDataAlgorithm’ was not declared in this scope; did you mean ‘cudnnGetConvolutionBackwardDataAlgorithm_v7’?
  866 |                 CHECK_CUDNN(cudnnGetConvolutionBackwardDataAlgorithm(
      |                             ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp:43:33: note: in definition of macro ‘CHECK_CUDNN’
   43 |     const cudnnStatus_t error = call;                                         \
      |                                 ^~~~
/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp:895:57: error: ‘CUDNN_CONVOLUTION_BWD_FILTER_PREFER_FASTEST’ was not declared in this scope; did you mean ‘CUDNN_CONVOLUTION_BWD_FILTER_ALGO_FFT’?
  895 |                         dnn_prefer_fastest_algorithms()?CUDNN_CONVOLUTION_BWD_FILTER_PREFER_FASTEST:CUDNN_CONVOLUTION_BWD_FILTER_NO_WORKSPACE,
      |                                                         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp:43:33: note: in definition of macro ‘CHECK_CUDNN’
   43 |     const cudnnStatus_t error = call;                                         \
      |                                 ^~~~
/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp:895:101: error: ‘CUDNN_CONVOLUTION_BWD_FILTER_NO_WORKSPACE’ was not declared in this scope; did you mean ‘CUDNN_CONVOLUTION_BWD_FILTER_ALGO_WINOGRAD’?
  895 |                         dnn_prefer_fastest_algorithms()?CUDNN_CONVOLUTION_BWD_FILTER_PREFER_FASTEST:CUDNN_CONVOLUTION_BWD_FILTER_NO_WORKSPACE,
      |                                                                                                     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp:43:33: note: in definition of macro ‘CHECK_CUDNN’
   43 |     const cudnnStatus_t error = call;                                         \
      |                                 ^~~~
/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp:889:29: error: ‘cudnnGetConvolutionBackwardFilterAlgorithm’ was not declared in this scope; did you mean ‘cudnnGetConvolutionBackwardFilterAlgorithm_v7’?
  889 |                 CHECK_CUDNN(cudnnGetConvolutionBackwardFilterAlgorithm(
      |                             ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp:43:33: note: in definition of macro ‘CHECK_CUDNN’
   43 |     const cudnnStatus_t error = call;                                         \
      |                                 ^~~~
make[2]: *** [CMakeFiles/cudnn_test.dir/build.make:63: CMakeFiles/cudnn_test.dir/home/alex/Build/dlib/dlib_git/dlib/cuda/cudnn_dlibapi.cpp.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:76: CMakeFiles/cudnn_test.dir/all] Error 2
make: *** [Makefile:84: all] Error 2
alex@ubsidian:~/Build/dlib/dlib_git/dlib/cmake_utils/test_for_cudnn/build$ 

Here's some information about my system:

alex@ubsidian:~$ uname -a
Linux ubsidian.malexmedia.net 5.4.0-37-generic #41-Ubuntu SMP Wed Jun 3 18:57:02 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux
alex@ubsidian:~$ 
alex@ubsidian:~$ cat /etc/os-release 
NAME="Ubuntu"
VERSION="20.04 LTS (Focal Fossa)"
ID=ubuntu
ID_LIKE=debian
PRETTY_NAME="Ubuntu 20.04 LTS"
VERSION_ID="20.04"
HOME_URL="https://www.ubuntu.com/"
SUPPORT_URL="https://help.ubuntu.com/"
BUG_REPORT_URL="https://bugs.launchpad.net/ubuntu/"
PRIVACY_POLICY_URL="https://www.ubuntu.com/legal/terms-and-policies/privacy-policy"
VERSION_CODENAME=focal
UBUNTU_CODENAME=focal
alex@ubsidian:~$ 
alex@ubsidian:~$ dpkg -l | grep -i cuda
ii  cuda                                       11.0.1-1                              amd64        CUDA meta-package
ii  cuda-11-0                                  11.0.1-1                              amd64        CUDA 11.0 meta-package
ii  cuda-command-line-tools-11-0               11.0.1-1                              amd64        CUDA command-line tools
ii  cuda-compiler-11-0                         11.0.1-1                              amd64        CUDA compiler
ii  cuda-cudart-11-0                           11.0.171-1                            amd64        CUDA Runtime native Libraries
ii  cuda-cudart-dev-11-0                       11.0.171-1                            amd64        CUDA Runtime native dev links, headers
ii  cuda-cuobjdump-11-0                        11.0.167-1                            amd64        CUDA cuobjdump
ii  cuda-cupti-11-0                            11.0.167-1                            amd64        CUDA profiling tools runtime libs.
ii  cuda-cupti-dev-11-0                        11.0.167-1                            amd64        CUDA profiling tools interface.
ii  cuda-demo-suite-11-0                       11.0.167-1                            amd64        Demo suite for CUDA
ii  cuda-documentation-11-0                    11.0.182-1                            amd64        CUDA documentation
ii  cuda-driver-dev-11-0                       11.0.171-1                            amd64        CUDA Driver native dev stub library
ii  cuda-drivers                               450.36.06-1                           amd64        CUDA Driver meta-package, branch-agnostic
ii  cuda-drivers-450                           450.36.06-1                           amd64        CUDA Driver meta-package, branch-specific
ii  cuda-gdb-11-0                              11.0.172-1                            amd64        CUDA-GDB
ii  cuda-libraries-11-0                        11.0.1-1                              amd64        CUDA Libraries 11.0 meta-package
ii  cuda-libraries-dev-11-0                    11.0.1-1                              amd64        CUDA Libraries 11.0 development meta-package
ii  cuda-memcheck-11-0                         11.0.167-1                            amd64        CUDA-MEMCHECK
ii  cuda-nsight-11-0                           11.0.167-1                            amd64        CUDA nsight
ii  cuda-nsight-compute-11-0                   11.0.1-1                              amd64        NVIDIA Nsight Compute
ii  cuda-nsight-systems-11-0                   11.0.1-1                              amd64        NVIDIA Nsight Systems
ii  cuda-nvcc-11-0                             11.0.167-1                            amd64        CUDA nvcc
ii  cuda-nvdisasm-11-0                         11.0.167-1                            amd64        CUDA disassembler
ii  cuda-nvml-dev-11-0                         11.0.167-1                            amd64        NVML native dev links, headers
ii  cuda-nvprof-11-0                           11.0.167-1                            amd64        CUDA Profiler tools
ii  cuda-nvprune-11-0                          11.0.167-1                            amd64        CUDA nvprune
ii  cuda-nvrtc-11-0                            11.0.167-1                            amd64        NVRTC native runtime libraries
ii  cuda-nvrtc-dev-11-0                        11.0.167-1                            amd64        NVRTC native dev links, headers
ii  cuda-nvtx-11-0                             11.0.167-1                            amd64        NVIDIA Tools Extension
ii  cuda-nvvp-11-0                             11.0.167-1                            amd64        CUDA Profiler tools
ii  cuda-runtime-11-0                          11.0.1-1                              amd64        CUDA Runtime 11.0 meta-package
ii  cuda-samples-11-0                          11.0.167-1                            amd64        CUDA example applications
ii  cuda-sanitizer-11-0                        11.0.167-1                            amd64        CUDA Sanitizer
ii  cuda-toolkit-11-0                          11.0.1-1                              amd64        CUDA Toolkit 11.0 meta-package
ii  cuda-tools-11-0                            11.0.1-1                              amd64        CUDA Tools meta-package
ii  cuda-visual-tools-11-0                     11.0.1-1                              amd64        CUDA visual tools
ii  libcudnn8                                  8.0.0.180-1+cuda11.0                  amd64        cuDNN runtime libraries
ii  libcudnn8-dev                              8.0.0.180-1+cuda11.0                  amd64        cuDNN development libraries and headers
ii  libcusolver-11-0                           10.4.0.191-1                          amd64        CUDA solver native runtime libraries
ii  libcusolver-dev-11-0                       10.4.0.191-1                          amd64        CUDA solver native dev links, headers
alex@ubsidian:~$ 

Note that, although Cuda 11 is a release candidate, and Ubuntu 20.04 is not technically supported, I have not had any other issues building Cuda projects with nvcc and Cuda is definitely working.

[alexmarkley]

Also of note, there appears to be no 7.6.x build of cuDNN for CUDA 11.

[rafale77]

exact same problem here as @alexmarkley on ubuntu 18.04LTS...

[davisking]

Pull the latest dlib from github and give it a try. I just updated it to support cuDNN 8.0.

[rafale77]

Just tested... almost there: It found it but somehow still says it won't use CUDA.

-- Using CMake version: 3.10.2
-- Compiling dlib version: 19.20.99
-- Enabling AVX instructions
-- Found system copy of libpng: /usr/lib/x86_64-linux-gnu/libpng.so;/usr/lib/x86_64-linux-gnu/libz.so
-- Found system copy of libjpeg: /usr/lib/x86_64-linux-gnu/libjpeg.so
-- Searching for BLAS and LAPACK
-- Searching for BLAS and LAPACK
-- Checking for module 'cblas'
--   No package 'cblas' found
-- Found Intel MKL BLAS/LAPACK library
-- Looking for cuDNN install...
-- Found cuDNN: /usr/lib/x86_64-linux-gnu/libcudnn.so
-- Disabling CUDA support for dlib.  DLIB WILL NOT USE CUDA
-- C++11 activated.
-- Configuring done
-- Generating done
-- Build files have been written to: /home/anhman/source/dlib/build

[davisking]

@rafale77 Are you building from a clean build folder?

[davisking]

Once cmake has detected cuDNN doesn't build, it won't check again. You have to clean out the build folder.

[rafale77]

Thanks! My bad...

Took me some time but I successfully finished compiling, installing the python API and am running it on home assistant now. The cnn model is indeed much faster. Watching a single stream from my doorbell consumes 2.5GB of memory though. I did not know this would consume this much GPU memory. I wonder if I should have picked a 1080Ti instead of a 2070...