I am trying to run darknet via the python binding inside a docker conainer.
It is running properly when I compile with GPU=0.
As base image, I used nvidia/cuda:10.2-base and copied the precompiled libdarknet.so (compiled with the same distro, 18.04) within the container.
nvidia-smi is running as well.
Once starting my script with the library compiled wth GPU=1 it crashes upon calling the from darknet import load_net with the following error:
OSError: libcublas.so.10: cannot open shared object file: No such file or directory
I tried using nvidia/cuda:10.0-base as base image as well, but then libcublas.so.10.2 is missing:
OSError: libcublas.so.10.2: cannot open shared object file: No such file or directory
Is it correct, that libdarknet.so is relying on both, CUDA 10 and CUDA 10.2?
Any tips on how to get it running with GPU enabled?
I am trying to run darknet via the python binding inside a docker conainer. It is running properly when I compile with
GPU=0
. As base image, I usednvidia/cuda:10.2-base
and copied the precompiledlibdarknet.so
(compiled with the same distro, 18.04) within the container.nvidia-smi
is running as well.Once starting my script with the library compiled wth
GPU=1
it crashes upon calling thefrom darknet import load_net
with the following error:OSError: libcublas.so.10: cannot open shared object file: No such file or directory
I tried using
nvidia/cuda:10.0-base
as base image as well, but thenlibcublas.so.10.2
is missing:OSError: libcublas.so.10.2: cannot open shared object file: No such file or directory
Is it correct, that libdarknet.so is relying on both, CUDA 10 and CUDA 10.2?
Any tips on how to get it running with GPU enabled?
Thanks a lot