train error: grad is being used in an in-place operation

[Sylvia-Vee]

Hi
excuse to ask you a question:

File "/home/czhan14x/train_project/ScaledYOLOv4/models/yolo.py", line 141, in _initialize_biases
b[:, 4] += math.log(8 / (640 / s) ** 2) # obj (8 objects per 640 image)
RuntimeError: a view of a leaf Variable that requires grad is being used in an in-place operation.

this error occurred while I train yolov4-large model in GeForce RTX 3090TI, but It's Ok in Titan RTX, I don't know why?
torch: 1.8.0+cu111
torchvision: 0.9.0+cu111

[FreeworkEarth]

Same error. Even on Colab

[FreeworkEarth]

Applying the follwing lines in the beginning of my Colab notebook fixed the problem, use the versions below and it should be able to train

!pip uninstall torch
!pip uninstall torch # run this command twice
!pip install torch==1.7.0+cu110 torchvision==0.8.1+cu110 torchtext==0.8 torchaudio===0.7.0 -f https://download.pytorch.org/whl/torch_stable.htmly

[Duarte-Nunes]

I gotta ask, is models/yolo.py even used in any other files of the repository?

[Sylvia-Vee]

I gotta ask, is models/yolo.py even used in any other files of the repository?

no, only this process

[Sylvia-Vee]

Applying the follwing lines in the beginning of my Colab notebook fixed the problem, use the versions below and it should be able to train

!pip uninstall torch
!pip uninstall torch # run this command twice
!pip install torch==1.7.0+cu110 torchvision==0.8.1+cu110 torchtext==0.8 torchaudio===0.7.0 -f https://download.pytorch.org/whl/torch_stable.htmly

Thanks, It works for me too