cause error when try to train

[jackyyang-cv]

the error is:RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.HalfTensor [1, 768]], which is output 0 of DivBackward0, is at version 1; expected version 0 instead. Hint: enable anomaly detection to find the operation that failed to compute its gradient, with torch.autograd.set_detect_anomaly(True).

[caoyunkang]

Hi,

Thanks for your interest! Your reported issue may be caused by a different version of Torch. I suggest downgrading your Torch to version 1.10.1+cu111. Best,

[jackyyang-cv]

I have fix this error.

[Hualiujie]

I have fix this error.

Do you downgrading your Torch to version 1.10.1+cu111? Can you share your method?

[Hualiujie]

Hi,

Thanks for your interest! Your reported issue may be caused by a different version of Torch. I suggest downgrading your Torch to version 1.10.1+cu111. Best,

I try to downgrading my Torch to version 1.10.1+cu111, but there are many errors. Who can share your method for this Error? Thank you

[Kaze816]

I have fix this error.

Do you downgrading your Torch to version 1.10.1+cu111? Can you share your method?

torch-1.13.1+cu116-cp39-cp39-win_amd64.whl wooking OK
torch-2.3.0+cu118-cp39-cp39-win_amd64.whl shows the same error
只需要切换虚拟环境即可

[alun011]

Hi,
Thanks for your interest! Your reported issue may be caused by a different version of Torch. I suggest downgrading your Torch to version 1.10.1+cu111. Best,

I try to downgrading my Torch to version 1.10.1+cu111, but there are many errors. Who can share your method for this Error? Thank you

hi, I have encountered the same problem as you, have you solved it? I have tried changing the torch version, but it has not been resolved

[caoyunkang]

hi, I have encountered the same problem as you, have you solved it? I have tried changing the torch version, but it has not been resolved

This issue primarily arises from in-place operations. Replacing /= and += with corresponding none-in-place operations in method/adaclip.py can resolve the problem.

[alun011]

hi, I have encountered the same problem as you, have you solved it? I have tried changing the torch version, but it has not been resolved

This issue primarily arises from in-place operations. Replacing /= and += with corresponding none-in-place operations in method/adaclip.py can resolve the problem.

Thank you! I solved my problem with your method.