Open varungupta31 opened 1 year ago
me too! Is that because we didn't install cudnn?
So I'm on a compute cluster for which I'm not the admin. However, I did end up locating the CUDNN files and fixed up the path in the Makefile, the error remains unsolved. I'm yet to understand exactly the role CUDNN=1
flag plays in the Makefile when CUDA=1
is already set. I built (make) the darknet using CUDA=1
and CUDNN=0
and it did compile. Try that and see if the GPU's are engaged.
Trying to make the darknet with the following Makefile with CUDA 10.2
As per the discussions here, I have updated the NVCC and the CUDA path under
LDFLAGS
, however runningmake
gives the following error:Please help out, I have checked out other relevant discussions on similar error stacks, to no resolve.