Run time error

[Emilio98HA]

When i run train.py it appears:
RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.FloatTensor [3, 12, 1, 1]] is at version 51; expected version 50 instead. Hint: enable anomaly detection to find the operation that failed to compute its gradient, with torch.autograd.set_detect_anomaly(True).

[nishavarghese15]

HI,

Did u solve this error? I also got the same error.

[Henry-GongZY]

HI,

Did u solve this error? I also got the same error.

You'd better use the version of python and pytorch mentioned in the repository. New version like pytorch 2.2 or pytorch 1.10 will not be ok.

[nishavarghese15]

Okay, Thanks..

[ChouMaoAn]

在model.py中
1.将 u_d, noise = self.denoise_net(u_list[-1])改为 u_d, noise = self.denoise_net(u_list[-1].detach())
2.将denoise_loss = self._denoise_criterion(u_list[-1], u_list[-2])改为denoise_loss = self._denoise_criterion(u_list[-1], u_list[-2].detach())

[XiongZhongxia]

在model.py中 1.将 u_d, noise = self.denoise_net(u_list[-1])改为 u_d, noise = self.denoise_net(u_list[-1].detach()) 2.将denoise_loss = self._denoise_criterion(u_list[-1], u_list[-2])改为denoise_loss = self._denoise_criterion(u_list[-1], u_list[-2].detach())

老哥好聪明啊，请问这是什么原理呢？

[zhuohai2363]

在model.py中 1.将 u_d, noise = self.denoise_net(u_list[-1])改为 u_d, noise = self.denoise_net(u_list[-1].detach()) 2.将denoise_loss = self._denoise_criterion(u_list[-1], u_list[-2])改为denoise_loss = self._denoise_criterion(u_list[-1], u_list[-2].detach())

我按照这个老哥的改了，但是还是没有跑起来，请问有其他人遇到和我一样的问题了嘛？