RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.FloatTensor [256, 50259]], which is output 0 of AsStridedBackward0, is at version 2; expected version 1 instead. Hint: enable anomaly detection to find the operation that failed to compute its gradient, with torch.autograd.set_detect_anomaly(True).

[ayssedgrmnci]

RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.FloatTensor [256, 50259]], which is output 0 of AsStridedBackward0, is at version 2; expected version 1 instead. Hint: enable anomaly detection to find the operation that failed to compute its gradient, with torch.autograd.set_detect_anomaly(True).

[rungjoo]

It seems to be similar to the this issue, have you followed it?
In the experimental environment torch>1.2, it is recommended to move the 129 line to the 143 line.

[ayssedgrmnci]

sorry i didn't see the other issue, thanks a lot for your help. The other thing I'm wondering is with which extension the trained model is registered and how can I run this model in the current time?

[rungjoo]

Link is the model inference code and link is the code that evaluates the model result.
Please refer to the above two codes for help.

You can refer to the paper for the model structure.
The result of the trained model is shown at link.