cuda driver version is insufficient for cuda runtime version.

[BlcaKHat]

!./darknet detector train data/obj.data cfg/yolo-obj.cfg darknet53.conv.74

CUDA Error: CUDA driver version is insufficient for CUDA runtime version
darknet: ./src/cuda.c:36: check_error: Assertion `0' failed.

i am trying on colab.

[derekwong66]

use cuda 9.0

[BlcaKHat]

I installed 9.2, is it not compatible ?

[AlexeyAB]

It isn't related to the Darknet.

Your CUDA version isn't compatible with you nVidia GPU Driver version: https://devtalk.nvidia.com/default/topic/1028320/cuda-driver-version-is-insufficient-for-cuda-runtime-version/
Update your nVidia GPU Driver, or downgrade your CUDA.
I recommend you to use CUDA 9.1.

[BlcaKHat]

git the same error with cuda 9.1

[BlcaKHat]

please help

[AlexeyAB]

Try to download the latest driver for your GPU: https://www.nvidia.ru/Download/index.aspx?lang=en

[BlcaKHat]

cuda 9.2 is latest as on nvidia website. I got the same error, then downgraded to 9.1 , got the same error.
I tried this same setup ofo code, and it was working well.
hard time.
can you check the cfg file once again.

yolo-obj-txt.txt

[AlexeyAB]

cfg-file looks good.

[BlcaKHat]

what would be the problem. ?
https://drive.google.com/open?id=1ugu5D3FKv7FDHdzycQuCrkAyS2vKmwE0
this is the drive link. if you get time please try running it .
it would be a great help

[ZYT5ting]

You should solve the CUDA driver problem first and then consider the data set. The error shows that your graphics card driver does not support the current version of CUDA, consider reducing the version of CUDA or updating your graphics card driver.

[ZYT5ting]

see solution1 or solution2

[BlcaKHat]

@ZYT5ting I was trying on google colabrotry, I tried with version 9.1 and 9.2.
both of them didn't worked.