CUDA driver version is insufficient for CUDA runtime version

AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )

http://pjreddie.com/darknet/

Other

21.71k stars 7.96k forks source link

CUDA driver version is insufficient for CUDA runtime version #3080

Open SouradipBh opened 5 years ago

SouradipBh commented 5 years ago

Hi, AlexeyAB, darknet.exe detector train data/obj.data cfg/yolo-obj.cfg darknet53.conv.74 At the time of running it, i faced a problem.It shows an error. The error is : CUDA status Error: file ....\src\dark_cuda.c : cuda_set_device() : line: 36 : build time: Apr 25 2019 - 12.07.50 CUDA Error: CUDA driver version is insufficient for CUDA runtime version

Can you please help me?

Error picture is : ERRROR

That is my NVSMI picture: NVSMI

That is my NVCC version :

NVCC VERSION

I have installed some software in my computer,as follows: OPENCV : 3.4.6-vc14_vc15 CUDA : 10.0.130_411.31_windows 10 cudnn : cudnn-10.0-windows10-x64-v7.4.2.24 Visual Studio:2015 NVIDIA GPU DRIVER : GEFORCE GT 635M - 2GB

xiphosuser commented 5 years ago

I’ve facing similar issue while trying to run Darknet. Having a GTX 1060 with cudnn 7.5. I’m also using cuda 10.0 with visual studio 2015.

Could someone please help?

Thanks

AlexeyAB commented 5 years ago

@SouradipBh @xiphosuser Hi,

It seems you should update your GPU driver.

Did you have another CUDA versions on your PC?
Try to set Install nVidia Driver during CUDA installation or install new GPU driver from: https://www.nvidia.ru/Download/Find.aspx?lang=en

SouradipBh commented 5 years ago

Thank You. @AlexeyAB

martisaju commented 5 years ago

@SouradipBh , @xiphosuser Hi,

Could you solve it? I am facing the same problem and I can't solve it... Thank you

SouradipBh commented 5 years ago

@martisaju This problem will solved in at least rtx 2080 ti graphics. I will suugest you to use darknet as a darknet_no_gpu.

vickyliuliu commented 4 years ago

Hello, I encountered the same issue. I have CUDA 10.1 and driver 435.21 on my computer, and I checked the official website of NVIDIA and found out that they should be compatible with each other. So I still cannot solve the problem... Any further suggestion? Thx