CUDA_ERROR_UNSUPPORTED_PTX_VERSION ptxas application ptx input, line 9; fatal : Unsupported .version 8.5; current version is '8.4'

[HadasRavikovitch]

Hi,
I have the same problem but this commands didnt solve it for me :(
Is there any other way you know?
Maybe something different in my case?

Originally posted by @HadasRavikovitch in #5081

[StevenGee3398]

I have a similar error when running my yolov4 darknet training code with cuda:

CUDA status Error: file: ./src/convolutional_kernels.cu: func: cuda_convert_f32_to_f16() line: 138

 CUDA Error: the provided PTX was compiled with an unsupported toolchain.
Darknet error location: ./src/convolutional_kernels.cu, cuda_convert_f32_to_f16(), line #138
CUDA Error: the provided PTX was compiled with an unsupported toolchain.: Success
backtrace (12 entries)

It used to work just fine before

[metrizable]

@HadasRavikovitch thanks for filing the issue and thanks for using Colab.

Can you provide a minimal reproducible notebook so that we can help troubleshoot?

[HadasRavikovitch]

Hi,
I got some help here.
These are the commandes that helped me:

!uv pip install -q --system --force-reinstall numba-cuda==0.4.0

from numba import config
config.CUDA_ENABLE_PYNVJITLINK = 1

Thank you

[HadasRavikovitch]

Hi,
Unfortunatly, the problem is back now.

Here is the reproducible notebook path -
https://github.com/HadasRavikovitch/Final-Project---GPU/blob/main/cross_time_tests_copy.ipynb

@metrizable Thank you!