dlib.DLIB_USE_CUDA.md

What to do when dlib.DLIB_USE_CUDA is False

FACE01 uses CUDA to maximize GPU utilization. Normally, by typing pip install dlib in the terminal, CUDA can be used according to the usage environment. To check if CUDA is available:

(FACE01) 
FACE01 $ pip freeze | grep dlib
dlib==19.24.0
(FACE01) 
FACE01 $ python
Python 3.8.10 (default, Nov 14 2022, 12:59:47) 
[GCC 9.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import dlib
>>> dlib.DLIB_USE_CUDA
True
>>> 

If False at this time, CUDA cannot be used.

Uninstall Dlib

Please uninstall Dlib once as follows.

pip uninstall dlib

Extract dlib-19.24.tar.bz2

Unzip dlib-19.24.tar.bz2 to create dlib-19.24 directory.

tar xvjf dlib-19.24.tar.bz2
cd dlib-19.24

Build with gcc-8

The gcc version is very important when building Dlib. gcc versions later than 8 are not supported. Specify gcc-8 etc. as follows.

(FACE01) 
FACE01/dlib-19.24 $ export CC=/usr/bin/gcc-8
(FACE01) 
FACE01/dlib-19.24 $ export CXX=/usr/bin/g++-8
(FACE01) 
FACE01/dlib-19.24 $ python setup.py install

Check if installed

(FACE01) 
FACE01/dlib-19.24 $ pip freeze | grep dlib
dlib==19.24.0
(FACE01) 
FACE01/dlib-19.24 $ python
Python 3.8.10 (default, Nov 14 2022, 12:59:47) 
[GCC 9.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import dlib
>>> dlib.DLIB_USE_CUDA
True
>>> 
(FACE01) 
FACE01/dlib-19.24 $ 

If you can confirm that it is True, you are finished:tada:.