You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I am a beginner in end-to-end autonomous driving and am trying to revive the InterFuser project.
My GPU is NVIDIA 4080 Super and it brings me some tough problems which hold me back.
I went through an extremely difficult experience and finally solved them, and now I wanna share it with you all.
If you find some problems in the solution or have better alternative please let me know. I'd appreciate it very much!
sm89 CUDA Capability
At first My CUDA version is 11.4 and Pytorch version is 1.12.x, virtual env with python 3.7 (created by requirements.txt)
BUT NVIDIA 40 Series's CUDA Capability is sm89, and the Traceback suggests that it is not compatible with the current PyTorch.
So some error emerge
After some research it suggests that sm89 needs CUDA 12.x
So I reinstall CUDA 12.1 and PyTorch which matches it, and find another problem.
PyTorch's oldest version which matches CUDA 12.x doesn't support Python 3.7
It at least need python 3.8
So I didn't listen to README.MD, created env with python 3.8
And this makes requirements.txt-installation meet some conflicts.
I made some adjustments by myself and try to run run_evaluation.sh
And I suspect this is why my timm module was not correctly installed.
No module named 'timm'
This is the third problem, when I run run_evaluation.sh
I first pip install timm, and clearly it didn't work.
Then I find some clue in Issues and find timm version was 0.4.13.
BUT pip install timm==0.4.13 didn't work too.
Just follow the instruction in the Setup section and use \interfuser\setup.py
Missing this may cause errors which doesn't note timm.
BTW, if running run_evaluation.sh with this result
just delete the file in InterFuser/results/sample_result.json
Hope that my experience can provide you with some references : )
The text was updated successfully, but these errors were encountered:
Thank you so much for your kind help! I’ve added the "good first issue" tag to attract more attention from newcomers and updated the README with the corresponding link. By the way, regarding the installation of timm, we’ve provided detailed instructions in the "Setup" section.
Thanks for reminding!For some reason, the installation was not successful at the beginning, and I carelessly ignored it later. I've modified origin issue.
I am a beginner in end-to-end autonomous driving and am trying to revive the InterFuser project.
My GPU is NVIDIA 4080 Super and it brings me some tough problems which hold me back.
I went through an extremely difficult experience and finally solved them, and now I wanna share it with you all.
If you find some problems in the solution or have better alternative please let me know. I'd appreciate it very much!
sm89
CUDA CapabilityAt first My CUDA version is 11.4 and Pytorch version is 1.12.x, virtual env with python 3.7 (created by requirements.txt)
BUT NVIDIA 40 Series's CUDA Capability is
sm89
, and the Traceback suggests that it is not compatible with the current PyTorch.So some error emerge
After some research it suggests that
sm89
needs CUDA 12.xSo I reinstall CUDA 12.1 and PyTorch which matches it, and find another problem.
PyTorch's oldest version which matches CUDA 12.x doesn't support Python 3.7
It at least need python 3.8
So I didn't listen to README.MD, created env with python 3.8
And this makes
requirements.txt
-installation meet some conflicts.I made some adjustments by myself and try to run
run_evaluation.sh
And I suspect this is why my
timm
module was not correctly installed.No module named 'timm'
This is the third problem, when I run
run_evaluation.sh
I first
pip install timm
, and clearly it didn't work.Then I find some clue in Issues and find timm version was 0.4.13.
BUT
pip install timm==0.4.13
didn't work too.Just follow the instruction in the Setup section and use
\interfuser\setup.py
Missing this may cause errors which doesn't note timm.
BTW, if running
run_evaluation.sh
with this resultjust delete the file in
InterFuser/results/sample_result.json
Hope that my experience can provide you with some references : )
The text was updated successfully, but these errors were encountered: