I have created a conda virtual env just to run dlib with gpu support, in it i installed cmake, CUDA and cuDNN from pip install cmake
and conda install cudatoolkit cudnn
, after that i tried to run dlibs setup.py from the site downloaded folder with python setup.py install --set DLIB_USE_CUDA=1 --set USE_AVX_INSTRUCTIONS=1
, but it didn't detect any cuda installation, as seem from the following section of the process:
Invoking CMake setup: 'cmake /home/gustavostahl/Downloads/dlib-19.19/tools/python -DCMAKE_LIBRARY_OUTPUT_DIRECTORY=/home/gustavostahl/Downloads/dlib-19.19/build/lib.linux-x86_64-3.7 -DPYTHON_EXECUTABLE=/home/gustavostahl/anaconda3/envs/dlib_gpu/bin/python -DDLIB_USE_CUDA=1 -DUSE_AVX_INSTRUCTIONS=1 -DCMAKE_BUILD_TYPE=Release'
-- pybind11 v2.2.2
-- Using CMake version: 3.16.3
-- Compiling dlib version: 19.19.0
-- SSE4 instructions can be executed by the host processor.
-- AVX instructions can be executed by the host processor.
-- Enabling AVX instructions
-- Searching for BLAS and LAPACK
-- Searching for BLAS and LAPACK
-- Checking for module 'cblas'
-- No package 'cblas' found
-- Found OpenBLAS library
-- Using OpenBLAS's built in LAPACK
-- Could NOT find CUDA: Found unsuitable version ".", but required is at least "7.5" (found /home/gustavostahl/anaconda3/envs/dlib_gpu/conda-meta)
-- DID NOT FIND CUDA
-- Disabling CUDA support for dlib. DLIB WILL NOT USE CUDA
-- C++11 activated.
My plan is to run dlib.train_shape_predictor
in the GPU to speed training time