darknet
darknet copied to clipboard
pjreddie
Fatal error: "cuda_runtime.h" not found
Hi, I am trying YOLO with CUDA. Following the guide reported in the website I change the Makefile enabling the flag GPU=1 but when I try to "make" the project again I get this error:
It is clear that there's a missing library (cuda_runtime.h). However, where am I supposed to add this library? At which path in my osx system? According to the control panel it seems that CUDA is correctly installed
I would really appreciate any help, thanks :-)
Run the following -
-
which nvcc You should see a path like /usr/local/cuda-8.0/bin/nvcc
-
Based on what path you see modify below command accordingly sudo ln -s /usr/local/cuda-8.0 /usr/local/cuda
This works in my implementation.
Same issue on Ubuntu 18.04, CUDA 9.0, cuDNN 7.14 with MakeFile only GPU = 1, all the else = 0.
Hello guys, my solution was to rollback to the previous version of nvidia cuda drivers: 387.10.10.10.30
Run the following -
- which nvcc You should see a path like /usr/local/cuda-8.0/bin/nvcc
- Based on what path you see modify below command accordingly sudo ln -s /usr/local/cuda-8.0 /usr/local/cuda
This works in my implementation.
does it really work?
@shaqing - Yes it does.. But be aware that CUDA 10 is the latest, so you need to create symlink of whatever version of Cuda you have..
so do:
ls /usr/local/cuda*
then create symbolic link to your version, for example, mine is "cuda-10.0" so you need to create symlink like this:
sudo ln -s /usr/local/cuda-10.0 /usr/local/cuda
When you run ln -s /usr/local/cuda-8.0 /usr/local/cuda can you explain what should be exactly these folders ? In my case I have nothing similar to both of them.
My CUDA_PATH is C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1 and when I run which nvcc I get /cygdrive/c/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/bin/nvcc
Someone could advise me please ? :)
@shaqing - Yes it does.. But be aware that CUDA 10 is the latest, so you need to create symlink of whatever version of Cuda you have..
so do:
ls /usr/local/cuda*then create symbolic link to your version, for example, mine is "cuda-10.0" so you need to create symlink like this:
sudo ln -s /usr/local/cuda-10.0 /usr/local/cuda
It's still not working too on my desktop bruh... look same
I am facing the same issue when I try to run "make" on windows 10. Did anyone find a solution to this?
@nijatmursali CUDA is for Nvidia, Apple since quite some time does not use Nvidia GPUs
